fix: Bedrock: Mantle models template api on ${AWS_REGION}, but aren't served in every region

models/alibaba/qwen3-vl-235b-a22b-instruct.toml

@@ -1,24 +0,0 @@
-name = "Qwen3 VL 235B A22B Instruct"
-description = "Qwen vision-language instruct model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-09-23"
-last_updated = "2025-09-23"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2025-03-31"
-open_weights = true
-
-[limit]
-context = 131_072
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Instruct"

models/alibaba/qwen3-vl-235b-a22b-thinking.toml

@@ -1,24 +0,0 @@
-name = "Qwen3 VL 235B A22B Thinking"
-description = "Qwen vision-language thinking model for visual reasoning, documents, and agent tasks"
-family = "qwen"
-release_date = "2025-09-23"
-last_updated = "2025-09-23"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2025-03-31"
-open_weights = true
-
-[limit]
-context = 131_072
-output = 32_768
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/Qwen/Qwen3-VL-235B-A22B-Thinking"

models/alibaba/qwen3.8-2.4t-a95b.toml

@@ -1,33 +0,0 @@
-# Sources (accessed 2026-08-16):
-# https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B
-# https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B/raw/main/README.md
-# https://qwen.ai/blog?id=qwen3.8
-# https://openrouter.ai/qwen/qwen3.8-2.4t-a95b
-# Open-weight twin of Qwen3.8 Max: text-only, thinking always on,
-# reasoning_effort low|medium|xhigh (default xhigh). Native context 262K,
-# extensible to ~1.01M. Distinct from closed multimodal qwen3.8-max.
-
-name = "Qwen3.8 2.4T A95B"
-description = "Open-weight sparse MoE (2.4T total, 95B active), the open-weight twin of Qwen3.8 Max for coding, research, complex reasoning, and agentic workflows"
-family = "qwen"
-release_date = "2026-08-12"
-last_updated = "2026-08-12"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = true
-license = "qwen3.8-max"
-
-[limit]
-context = 262_144
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B"

models/deepseek/deepseek-v4-flash-vision-exp.toml

@@ -1,19 +0,0 @@
-name = "DeepSeek V4 Flash Vision Exp"
-description = "Experimental multimodal DeepSeek V4 Flash model for image understanding, coding, and agentic work"
-family = "deepseek-flash"
-release_date = "2026-08-21"
-last_updated = "2026-08-21"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[limit]
-context = 1_000_000
-output = 384_000
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

models/deepseek/deepseek-v4-pro-0423.toml

@@ -1,22 +0,0 @@
-# https://ofox.ai/models/deepseek/deepseek-v4-pro-0423
-# https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro
-name = "DeepSeek V4 Pro 0423"
-description = "DeepSeek V4 Pro initial snapshot with million-token context and support for thinking and non-thinking modes"
-family = "deepseek-thinking"
-release_date = "2026-04-23"
-last_updated = "2026-04-23"
-attachment = false
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-knowledge = "2025-05"
-open_weights = true
-
-[limit]
-context = 1_000_000
-output = 384_000
-
-[modalities]
-input = ["text"]
-output = ["text"]

models/deepseek/deepseek-v4-pro-0813.toml

@@ -1,16 +1,14 @@
-# https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813
 name = "DeepSeek V4 Pro 0813"
 description = "DeepSeek V4 Pro snapshot with million-token context and support for thinking and non-thinking modes"
 family = "deepseek-thinking"
 release_date = "2026-08-12"
-last_updated = "2026-08-22"
+last_updated = "2026-08-12"
 attachment = false
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-open_weights = true
-license = "MIT"
+open_weights = false
 
 [limit]
 context = 1_000_000
@@ -19,7 +17,3 @@ output = 384_000
 [modalities]
 input = ["text"]
 output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro-0813"

models/google/gemini-flash-latest.toml

@@ -1,15 +1,15 @@
-# Tracks the current Gemini Flash release (gemini-3.7-flash).
+# Tracks the current Gemini Flash release (gemini-3.5-flash).
 name = "Gemini Flash Latest"
-description = "High-efficiency Gemini model for agentic workflows, coding, and multimodal reasoning"
+description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
 family = "gemini-flash"
-release_date = "2026-08-13"
-last_updated = "2026-08-13"
+release_date = "2026-05-19"
+last_updated = "2026-05-19"
 attachment = true
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-knowledge = "2026-03"
+knowledge = "2025-01"
 open_weights = false
 
 [limit]

models/google/gemini-flash-lite-latest.toml

@@ -1,15 +1,15 @@
-# Tracks the current Gemini Flash-Lite release (gemini-3.5-flash-lite).
+# Tracks the current Gemini Flash-Lite release (gemini-3.1-flash-lite).
 name = "Gemini Flash-Lite Latest"
-description = "Fast Gemini model balancing multimodal reasoning, tool use, and cost"
+description = "Low-latency Gemini model for high-volume multimodal and agent workloads"
 family = "gemini-flash-lite"
-release_date = "2026-07-21"
-last_updated = "2026-07-21"
+release_date = "2026-05-07"
+last_updated = "2026-05-07"
 attachment = true
 reasoning = true
 temperature = true
 tool_call = true
 structured_output = true
-knowledge = "2026-03"
+knowledge = "2025-01"
 open_weights = false
 
 [limit]

models/mistral/voxtral-small-latest.toml

@@ -1,24 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://docs.mistral.ai/models/model-cards/voxtral-small-25-07
-#   https://mistral.ai/news/voxtral/
-# Field values mirror Mistral's own first-party host entry in this repo
-# (providers/mistral/models/voxtral-small-latest.toml); host-scoped keys
-# (cost, status) are intentionally left to the provider files.
-name = "Voxtral Small (latest)"
-description = "Instruct model with native audio input for speech understanding and tool use"
-family = "voxtral"
-release_date = "2025-07-15"
-last_updated = "2025-07-15"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = true
-
-[limit]
-context = 32_000
-output = 32_000
-
-[modalities]
-input = ["text", "audio"]
-output = ["text"]

models/perplexity/sonar-deep-research.toml

@@ -1,25 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://docs.perplexity.ai/docs/sonar/models/sonar-deep-research
-#   https://docs.perplexity.ai/api-reference/sonar-post
-# Field values mirror Perplexity's own first-party host entry in this repo
-# (providers/perplexity/models/sonar-deep-research.toml); host-scoped keys
-# (cost, reasoning_options) are intentionally left to the provider files.
-name = "Sonar Deep Research"
-description = "Sonar search model for autonomous research and citation-backed long-form reports"
-family = "sonar"
-release_date = "2025-02-01"
-last_updated = "2025-09-01"
-attachment = false
-reasoning = true
-temperature = false
-tool_call = false
-knowledge = "2025-01"
-open_weights = false
-
-[limit]
-context = 128_000
-output = 32_768
-
-[modalities]
-input = ["text"]
-output = ["text"]

models/sakana/sakana-namazu.toml

@@ -1,59 +0,0 @@
-name = "Sakana Namazu"
-description = "Japanese-specialized reasoning model based on Kimi K2.6 and tuned for Japanese language, culture, and business workflows"
-family = "sakana-namazu"
-release_date = "2026-08-03"
-last_updated = "2026-08-03"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-structured_output = true
-open_weights = false
-
-[limit]
-context = 262_144
-output = 65_536
-
-[modalities]
-input = ["text", "image", "pdf"]
-output = ["text"]
-
-[[links]]
-label = "Official product page"
-url = "https://sakana.ai/namazu/"
-type = "announcement"
-
-[[links]]
-label = "Official model documentation"
-url = "https://console.sakana.ai/models?model=sakana-namazu"
-type = "docs"
-
-[[benchmarks]]
-name = "AIME26"
-score = 96.67
-source = "https://console.sakana.ai/models?model=sakana-namazu"
-
-[[benchmarks]]
-name = "MMLU-Pro"
-score = 90.33
-source = "https://console.sakana.ai/models?model=sakana-namazu"
-
-[[benchmarks]]
-name = "LiveCodeBench v6"
-score = 90.33
-source = "https://console.sakana.ai/models?model=sakana-namazu"
-
-[[benchmarks]]
-name = "JFBench"
-score = 37.40
-source = "https://console.sakana.ai/models?model=sakana-namazu"
-
-[[benchmarks]]
-name = "Translation"
-score = 52.20
-source = "https://console.sakana.ai/models?model=sakana-namazu"
-
-[[benchmarks]]
-name = "FairPoliticsQA"
-score = 56.30
-source = "https://console.sakana.ai/models?model=sakana-namazu"

models/swiss-ai/apertus-8b.toml

@@ -1,33 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://huggingface.co/swiss-ai/Apertus-8B-Instruct-2509
-#   https://arxiv.org/abs/2509.14233
-# The model card states "Apertus by default supports a context length up to 65,536 tokens",
-# Apache-2.0 licensing, and tool use support. Sibling entry: models/swiss-ai/apertus-70b.toml.
-name = "Apertus 8B"
-description = "Fully open 8B multilingual LLM supporting 1800+ languages with 65K context. Trained on compliant open data. Apache 2.0, EU AI Act compliant."
-release_date = "2025-09-02"
-last_updated = "2025-09-02"
-knowledge = "2025-09"
-attachment = false
-reasoning = false
-temperature = true
-tool_call = true
-open_weights = true
-license = "Apache-2.0"
-
-[limit]
-context = 65_536
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/swiss-ai/Apertus-8B-Instruct-2509"
-
-[[links]]
-label = "Paper"
-url = "https://arxiv.org/abs/2509.14233"
-type = "paper"

models/trendyol/asure-12b.toml

@@ -1,30 +0,0 @@
-# Sources:
-# https://huggingface.co/Trendyol/Trendyol-LLM-Asure-12B
-# https://huggingface.co/api/models/Trendyol/Trendyol-LLM-Asure-12B (createdAt, license, base_model)
-# https://huggingface.co/Trendyol/Trendyol-LLM-Asure-12B/raw/main/config.json (max_position_embeddings)
-# `reasoning` and `tool_call` are not stated on the model card; both were
-# measured against a host serving these weights (llmtr.com, 2026-08-16):
-# a request carrying `tools` returns no tool_calls, and no reasoning output
-# is produced.
-name = "Trendyol Asure 12B"
-description = "Turkish-language multimodal instruct model built on Gemma 3 12B for e-commerce text, chat, and image-text tasks"
-family = "gemma"
-release_date = "2026-02-19"
-last_updated = "2026-02-20"
-attachment = true
-reasoning = false
-temperature = true
-tool_call = false
-open_weights = true
-license = "Gemma"
-
-[limit]
-context = 131_072
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/Trendyol/Trendyol-LLM-Asure-12B"

models/upstage/solar-pro2.toml

@@ -1,25 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://developers.upstage.ai/docs/apis/chat
-#   https://developers.upstage.ai/docs/capabilities/generate/reasoning
-# Field values mirror Upstage's own first-party host entry in this repo
-# (providers/upstage/models/solar-pro2.toml); host-scoped keys (cost,
-# reasoning_options) are intentionally left to the provider files.
-name = "Solar Pro 2"
-description = "Flagship model for demanding analysis, coding, and production agent workflows"
-family = "solar-pro"
-release_date = "2025-05-20"
-last_updated = "2025-05-20"
-attachment = false
-reasoning = true
-temperature = true
-knowledge = "2025-03"
-tool_call = true
-open_weights = false
-
-[limit]
-context = 65_536
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]

models/upstage/solar-pro3.toml

@@ -1,25 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://developers.upstage.ai/docs/apis/chat
-#   https://developers.upstage.ai/docs/capabilities/generate/reasoning
-# Field values mirror Upstage's own first-party host entry in this repo
-# (providers/upstage/models/solar-pro3.toml); host-scoped keys (cost,
-# reasoning_options) are intentionally left to the provider files.
-name = "Solar Pro 3"
-description = "Flagship model for demanding analysis, coding, and production agent workflows"
-family = "solar-pro"
-release_date = "2026-01"
-last_updated = "2026-01"
-attachment = false
-reasoning = true
-temperature = true
-knowledge = "2025-03"
-tool_call = true
-open_weights = false
-
-[limit]
-context = 131_072
-output = 8_192
-
-[modalities]
-input = ["text"]
-output = ["text"]

models/upstage/solar-pro4.toml

@@ -1,26 +0,0 @@
-# Sources (accessed 2026-08-16):
-#   https://developers.upstage.ai/docs/apis/chat
-#   https://developers.upstage.ai/docs/capabilities/generate/reasoning
-# Field values mirror Upstage's own first-party host entry in this repo
-# (providers/upstage/models/solar-pro4.toml); host-scoped keys (cost,
-# reasoning_options) are intentionally left to the provider files.
-name = "Solar Pro 4"
-description = "Upstage's flagship model, specialized for agentic use"
-family = "solar-pro"
-release_date = "2026-08-06"
-last_updated = "2026-08-06"
-attachment = false
-reasoning = true
-temperature = true
-knowledge = "2026-02"
-tool_call = true
-structured_output = true
-open_weights = false
-
-[limit]
-context = 524_288
-output = 131_072
-
-[modalities]
-input = ["text"]
-output = ["text"]

models/xai/grok-imagine-image-2.0.toml

@@ -1,26 +0,0 @@
-# Sources:
-# - https://docs.x.ai/docs/models
-# - https://docs.x.ai/developers/models/grok-imagine-image-2.0
-# - https://docs.x.ai/docs/guides/image-generation
-# - https://x.ai/news/grok-imagine-image-2
-# Pricing: $0.04 per image (not token-based; no [cost] authored)
-# Release: 2026-08-07 (GA as Quality Mode; API model id grok-imagine-image-2.0)
-
-name = "Grok Imagine Image 2.0"
-description = "Image model for prompt-driven generation, editing, and visual design workflows"
-family = "grok"
-release_date = "2026-08-07"
-last_updated = "2026-08-07"
-attachment = true
-reasoning = false
-temperature = false
-tool_call = false
-open_weights = false
-
-[limit]
-context = 8_000
-output = 0
-
-[modalities]
-input = ["text", "image"]
-output = ["image"]

models/zhipuai/glm-4.6v-flash.toml

@@ -1,25 +0,0 @@
-# Sources (accessed 2026-08-19):
-# - https://z.ai/blog/glm-4.6v
-# - https://huggingface.co/zai-org/GLM-4.6V-Flash
-name = "GLM-4.6V-Flash"
-description = "Lightweight GLM vision model for visual reasoning, documents, and multimodal agents"
-family = "glm"
-release_date = "2025-12-08"
-last_updated = "2025-12-08"
-attachment = true
-reasoning = true
-temperature = true
-tool_call = true
-open_weights = true
-
-[limit]
-context = 128_000
-output = 32_768
-
-[modalities]
-input = ["text", "image", "video"]
-output = ["text"]
-
-[[weights]]
-label = "Hugging Face"
-url = "https://huggingface.co/zai-org/GLM-4.6V-Flash"

package.json

@@ -28,7 +28,6 @@
     "huggingface:sync": "bun ./packages/core/script/sync-models.ts huggingface",
     "kilo:sync": "bun ./packages/core/script/sync-models.ts kilo",
     "llmgateway:sync": "bun ./packages/core/script/sync-models.ts llmgateway",
-    "llmgateway-providers:sync": "bun ./packages/core/script/sync-models.ts llmgateway-providers",
     "requesty:sync": "bun ./packages/core/script/sync-models.ts requesty",
     "merge-gateway:sync": "bun ./packages/core/script/sync-models.ts merge-gateway",
     "nano-gpt:sync": "bun ./packages/core/script/sync-models.ts nano-gpt",

packages/core/src/family.ts

@@ -52,7 +52,6 @@ export const ModelFamilyValues = [
 
   // Meta Muse
   "muse",
-  "muse-free",
 
   // Alibaba Qwen
   "qwen",
@@ -367,9 +366,6 @@ export const ModelFamilyValues = [
   // Conductor
   "fugu",
 
-  // Sakana Namazu
-  "sakana-namazu",
-
   // V0
   "v0",
 

packages/core/src/sync/auto-merge.ts

@@ -5,13 +5,9 @@ export const MAX_CREATED_MODELS = 10;
 export const MAX_DELETED_MODELS = 10;
 export const MAX_MODEL_CHURN = 15;
 const REVIEWED_REASONING_PROVIDERS = new Set([
-  "crossmodel",
-  "edenai",
   "empiriolabs",
-  "hyper",
   "kilo",
   "llmgateway",
-  "llmgateway-providers",
   "merge-gateway",
   "nano-gpt",
   "openrouter",

packages/core/src/sync/index.ts

@@ -21,7 +21,7 @@ import { hyper } from "./providers/hyper.js";
 import { huggingface } from "./providers/huggingface.js";
 import { inceptron } from "./providers/inceptron.js";
 import { kilo } from "./providers/kilo.js";
-import { llmgateway, llmgatewayProviders } from "./providers/llmgateway.js";
+import { llmgateway } from "./providers/llmgateway.js";
 import { mergeGateway } from "./providers/merge-gateway.js";
 import { nanoGpt } from "./providers/nano-gpt.js";
 import { openai } from "./providers/openai.js";
@@ -98,17 +98,7 @@ export interface SyncProvider<SourceModel> {
       existing(id: string): ExistingModel | undefined;
       authored(id: string): ExistingModel | undefined;
     },
-  ): {
-    id: string;
-    model: SyncedModel;
-    metadata?: { id: string; model: SyncedMetadata };
-    /**
-     * Leading comment block for the written file when it has none of its own
-     * (e.g. the wire-path header every toggle reasoning control requires). A
-     * header already present on the existing file always wins.
-     */
-    header?: string;
-  } | undefined;
+  ): { id: string; model: SyncedModel; metadata?: { id: string; model: SyncedMetadata } } | undefined;
 }
 
 export interface SyncResult {
@@ -141,7 +131,6 @@ export const providers: {
   inceptron: SyncProvider<any>;
   kilo: SyncProvider<any>;
   llmgateway: SyncProvider<any>;
-  "llmgateway-providers": SyncProvider<any>;
   "merge-gateway": SyncProvider<any>;
   "nano-gpt": SyncProvider<any>;
   ofox: SyncProvider<any>;
@@ -173,7 +162,6 @@ export const providers: {
   inceptron,
   kilo,
   llmgateway,
-  "llmgateway-providers": llmgatewayProviders,
   "merge-gateway": mergeGateway,
   "nano-gpt": nanoGpt,
   ofox,
@@ -198,7 +186,6 @@ export const groups = {
     "inceptron",
     "kilo",
     "llmgateway",
-    "llmgateway-providers",
     "merge-gateway",
     "nano-gpt",
     "ofox",
@@ -283,16 +270,13 @@ export async function syncProvider<SourceModel>(
       : preserveBaseModel(translated.model, existing.get(relativePath)?.authored);
     const translatedBase = "base_model" in translatedModel ? translatedModel.base_model : undefined;
     let resolvedReasoning: boolean | undefined;
-    let baseReasoningOptions: unknown;
     if (translatedBase !== undefined) {
       if (translated.metadata?.id === translatedBase) {
         resolvedReasoning = translated.metadata.model.reasoning;
-        baseReasoningOptions = translated.metadata.model.reasoning_options;
       } else {
         modelMetadata ??= await readModelMetadata(provider.modelsDir);
         const canonicalReasoning = modelMetadata[translatedBase]?.reasoning;
         resolvedReasoning = typeof canonicalReasoning === "boolean" ? canonicalReasoning : undefined;
-        baseReasoningOptions = modelMetadata[translatedBase]?.reasoning_options;
       }
     } else {
       resolvedReasoning = existing.get(relativePath)?.toml.reasoning;
@@ -301,7 +285,6 @@ export async function syncProvider<SourceModel>(
       translatedModel,
       existing.get(relativePath)?.authored,
       resolvedReasoning,
-      baseReasoningOptions,
     );
     const withDescription = provider.preserveDescriptions === false
       ? withReasoningOptions
@@ -317,7 +300,7 @@ export async function syncProvider<SourceModel>(
 
     desired.set(relativePath, {
       model: parsed.data,
-      content: ((existing.get(relativePath)?.header || translated.header) ?? "") + formatToml(parsed.data),
+      content: (existing.get(relativePath)?.header ?? "") + formatToml(parsed.data),
     });
   }
 
@@ -493,7 +476,6 @@ export function preserveReasoningOptions(
   model: SyncedModel,
   existing: ExistingModel | undefined,
   resolvedReasoning: boolean | undefined = existing?.reasoning,
-  baseReasoningOptions: unknown = undefined,
 ): SyncedModel {
   if ((model.reasoning ?? resolvedReasoning) === false) {
     const { reasoning_options: _reasoningOptions, ...withoutReasoningOptions } = model;
@@ -501,10 +483,7 @@ export function preserveReasoningOptions(
   }
   if (model.reasoning_options !== undefined) return model;
   if (existing?.reasoning_options === undefined) {
-    // When the base model already declares reasoning_options, leave the field
-    // unset so the factored file inherits them — stamping [] here would
-    // shadow the base's real controls with "no controls".
-    return (model.reasoning ?? resolvedReasoning) === true && baseReasoningOptions === undefined
+    return (model.reasoning ?? resolvedReasoning) === true
       ? { ...model, reasoning_options: [] }
       : model;
   }

packages/core/src/sync/providers/baseten.ts

@@ -1,13 +1,7 @@
 import { z } from "zod";
 
 import { describeModel } from "../../describe.js";
-import type {
-  ExistingModel,
-  SyncProvider,
-  SyncedBaseModel,
-  SyncedFullModel,
-  SyncedModel,
-} from "../index.js";
+import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
 import { factorBaseModel, resolveCanonicalBaseModel } from "./openrouter.js";
 
 const API_ENDPOINT = "https://inference.baseten.co/v1/models";
@@ -67,7 +61,6 @@ export const baseten = {
   },
   translateModel(model, context) {
     const existing = context.existing(model.id);
-    const authored = context.authored(model.id);
     const baseModel = existing === undefined
       ? resolveBasetenBaseModel(model.id)
       : existing.base_model;
@@ -79,7 +72,7 @@ export const baseten = {
 
     return {
       id: model.id,
-      model: buildBasetenModel(model, existing, baseModel, authored),
+      model: buildBasetenModel(model, existing, baseModel),
     };
   },
 } satisfies SyncProvider<BasetenModel>;
@@ -109,7 +102,6 @@ export function buildBasetenModel(
   model: BasetenModel,
   existing: ExistingModel | undefined,
   baseModel = existing === undefined ? resolveBasetenBaseModel(model.id) : existing.base_model,
-  authored?: ExistingModel,
 ): SyncedModel {
   const features = new Set(model.supported_features);
   const samplingParameters = new Set(model.supported_sampling_parameters);
@@ -130,9 +122,7 @@ export function buildBasetenModel(
   const limit = {
     context: model.context_length,
     input: existing?.limit?.input,
-    // Explicit provider limits are serving overrides and sync pins. Baseten's
-    // catalog has returned a model's context window as its completion limit.
-    output: authored?.limit?.output ?? model.max_completion_tokens,
+    output: model.max_completion_tokens,
   };
   const values: Partial<SyncedFullModel> = {
     name: model.name ?? existing?.name,
@@ -169,21 +159,7 @@ export function buildBasetenModel(
     if (limit.context === undefined || limit.output === undefined) {
       throw new Error(`Baseten model ${model.id} has incomplete token limits required for sync`);
     }
-    const factored = factorBaseModel(
-      baseModel,
-      values,
-      limit,
-      existing?.base_model_omit,
-    ) as SyncedBaseModel;
-    if (authored?.limit?.output === undefined) return factored;
-
-    return {
-      ...factored,
-      limit: {
-        ...factored.limit,
-        output: authored.limit.output,
-      },
-    };
+    return factorBaseModel(baseModel, values, limit, existing?.base_model_omit);
   }
 
   const required = z.object({

packages/core/src/sync/providers/cortecs.ts

@@ -13,16 +13,6 @@ const CANONICAL_BASE_MODEL_EXCEPTIONS = {
 const EUR_TO_USD = 1.114;
 
 const CortecsModality = z.enum(["text", "audio", "image", "video", "pdf"]);
-type CortecsModality = z.infer<typeof CortecsModality>;
-
-function modalities(values: string[]): CortecsModality[] {
-  const allowed = new Set<CortecsModality>(CortecsModality.options);
-  const result = values
-    .map((value) => value.toLowerCase())
-    .map((value) => value === "file" ? "pdf" : value)
-    .filter((value): value is CortecsModality => allowed.has(value as CortecsModality));
-  return [...new Set<CortecsModality>(result.length > 0 ? result : ["text"])];
-}
 
 export const CortecsModel = z.object({
   id: z.string().min(1),
@@ -36,8 +26,8 @@ export const CortecsModel = z.object({
     cache_write_cost: z.number().nonnegative().optional(),
   }).passthrough(),
   context_size: z.number().int().positive(),
-  input_modalities: z.array(z.string()).transform(modalities).default(["text"]),
-  output_modalities: z.array(z.string()).transform(modalities).default(["text"]),
+  input_modalities: z.array(CortecsModality).default(["text"]),
+  output_modalities: z.array(CortecsModality).default(["text"]),
   supported_features: z.array(z.string()).default([]),
 }).passthrough();
 

packages/core/src/sync/providers/crossmodel.ts

@@ -22,13 +22,11 @@ function baseModelExists(modelID: string): boolean {
 // CROSSMODEL_MODELS_URL overrides the endpoint (e.g. a local backend) for testing.
 const API_ENDPOINT = process.env.CROSSMODEL_MODELS_URL ?? "https://www.crossmodel.ai/api/models";
 
-const REASONING_EFFORTS = ["none", "minimal", "low", "medium", "high", "xhigh", "max", "default"] as const;
-
 const ReasoningCapability = z
   .object({
     supported: z.boolean().optional(),
     toggle: z.boolean().nullish().transform((value) => value ?? undefined),
-    effort: z.array(z.enum(REASONING_EFFORTS)).nullish().transform((value) => value ?? undefined),
+    effort: z.array(z.string()).nullish().transform((value) => value ?? undefined),
     budget_tokens: z
       .object({ min: z.number().optional(), max: z.number().optional() })
       .nullish()
@@ -162,6 +160,15 @@ function modalities(values: string[] | undefined, fallback: Modality[]): Modalit
   return [...new Set(result.length > 0 ? result : fallback)];
 }
 
+// models.dev's reasoning_options effort enum (schema.ts ReasoningEffortValue).
+// Guarding against it means an unexpected upstream value is dropped instead of
+// silently producing a TOML that fails `validate`.
+const REASONING_EFFORTS = ["none", "minimal", "low", "medium", "high", "xhigh", "max", "default"] as const;
+type ReasoningEffort = (typeof REASONING_EFFORTS)[number];
+function isReasoningEffort(value: string): value is ReasoningEffort {
+  return (REASONING_EFFORTS as readonly string[]).includes(value);
+}
+
 // Project CrossModel's capabilities.reasoning onto models.dev reasoning_options.
 //   reasoning absent  -> undefined (non-reasoning model; option omitted)
 //   reasoning === {}  -> [] (model reasons, no verified user-selectable control)
@@ -172,7 +179,8 @@ function reasoningOptions(model: CrossModelModel): SyncedModel["reasoning_option
   const options: NonNullable<SyncedModel["reasoning_options"]> = [];
   if (reasoning.toggle === true) options.push({ type: "toggle" });
   if (reasoning.effort !== undefined) {
-    if (reasoning.effort.length > 0) options.push({ type: "effort", values: reasoning.effort });
+    const values = reasoning.effort.filter(isReasoningEffort);
+    if (values.length > 0) options.push({ type: "effort", values });
   }
   if (reasoning.budget_tokens !== undefined) {
     const budget: { type: "budget_tokens"; min?: number; max?: number } = { type: "budget_tokens" };

packages/core/src/sync/providers/llmgateway.ts

@@ -1,21 +1,16 @@
 import { z } from "zod";
-import { existsSync, readFileSync } from "node:fs";
-import path from "node:path";
 
 import { describeModel } from "../../describe.js";
 import { inferKimiFamily, ModelFamilyValues } from "../../family.js";
 import { ReasoningOption } from "../../schema.js";
 import type { ExistingModel, SyncProvider, SyncedFullModel, SyncedModel } from "../index.js";
-import { factorBaseModel, resolveModelMetadataBaseModel } from "./openrouter.js";
+import { factorBaseModel, resolveCanonicalBaseModel } from "./openrouter.js";
 
 const API_ENDPOINT = "https://api.llmgateway.io/v1/models";
 
 // LLM Gateway names the originating lab in `family`; most already match the
-// canonical prefixes understood by resolveModelMetadataBaseModel, and labs
-// outside that shared table (e.g. perplexity) resolve through its exact
-// `models/` path match without widening the OpenRouter prefix map for every
-// other provider. Alias the few that spell the lab differently. (Mirrors
-// huggingface's CANONICAL_ORG_PREFIXES.)
+// canonical prefixes understood by resolveCanonicalBaseModel. Alias the few that
+// spell the lab differently. (Mirrors huggingface's CANONICAL_ORG_PREFIXES.)
 const CANONICAL_FAMILY_ALIASES: Record<string, string> = {
   grok: "xai",
   mistral: "mistralai",
@@ -35,6 +30,10 @@ const Pricing = z.object({
   input_cache_write: z.string().optional(),
 });
 
+const LLMGatewayProvider = z.object({
+  reasoning_efforts: z.array(z.string()).optional(),
+}).passthrough();
+
 const ReasoningEffortOrder = new Map([
   "none",
   "minimal",
@@ -55,22 +54,9 @@ export const LLMGatewayModel = z.object({
     input_modalities: z.array(z.string()),
     output_modalities: z.array(z.string()),
   }),
-  providers: z.array(
-    z.object({
-      providerId: z.string().optional(),
-      vision: z.boolean().optional(),
-      tools: z.boolean().optional(),
-      reasoning: z.boolean().optional(),
-      reasoning_efforts: z.array(
-        z.enum(["none", "minimal", "low", "medium", "high", "xhigh", "max", "default"]),
-      ).optional(),
-    }).passthrough(),
-  ).optional(),
   pricing: Pricing,
-  // Absent for pseudo-models (custom/auto) and some non-text mappings; text
-  // models always report it.
-  context_length: z.number().optional(),
-  max_output: z.number().optional(),
+  providers: z.array(LLMGatewayProvider),
+  context_length: z.number(),
   supported_parameters: z.array(z.string()),
   structured_outputs: z.boolean().optional(),
 }).passthrough();
@@ -81,117 +67,31 @@ export const LLMGatewayResponse = z.object({
 
 export type LLMGatewayModel = z.infer<typeof LLMGatewayModel>;
 
-async function fetchLLMGatewayModels(url: string) {
-  const headers = process.env.LLMGATEWAY_API_KEY
-    ? { Authorization: `Bearer ${process.env.LLMGATEWAY_API_KEY}` }
-    : undefined;
-  const response = await fetch(url, { headers });
-  if (!response.ok) {
-    throw new Error(`LLM Gateway request failed: ${response.status} ${response.statusText}`);
-  }
-  return response.json();
-}
-
-function textOnly(model: LLMGatewayModel) {
-  const output = model.architecture.output_modalities;
-  return output.length === 1 && output[0] === "text";
-}
-
-// The DevPass (LLM Gateway) provider: the gateway's aggregated catalog of root
-// model IDs, auto-routed across upstream providers.
 export const llmgateway = {
   id: "llmgateway",
-  name: "DevPass (LLM Gateway)",
+  name: "LLM Gateway",
   modelsDir: "providers/llmgateway/models",
   async fetchModels() {
-    return fetchLLMGatewayModels(API_ENDPOINT);
+    const headers = process.env.LLMGATEWAY_API_KEY
+      ? { Authorization: `Bearer ${process.env.LLMGATEWAY_API_KEY}` }
+      : undefined;
+    const response = await fetch(API_ENDPOINT, { headers });
+    if (!response.ok) {
+      throw new Error(`LLM Gateway request failed: ${response.status} ${response.statusText}`);
+    }
+    return response.json();
   },
   parseModels(raw) {
-    const data = LLMGatewayResponse.parse(raw).data.filter(textOnly);
-    // An empty catalog is an upstream fault; syncing it would delete every
-    // model file, so fail loudly instead.
-    if (data.length === 0) {
-      throw new Error("LLM Gateway returned no text models");
-    }
-    return data;
+    return LLMGatewayResponse.parse(raw).data.filter((model) => {
+      const output = model.architecture.output_modalities;
+      return output.length === 1 && output[0] === "text";
+    });
   },
   translateModel(model, context) {
-    const translated = buildLLMGatewayModel(model, context.existing(model.id));
-    if (translated === undefined) {
-      return undefined;
-    }
-    return { id: model.id, model: translated };
-  },
-  sourceID(model) {
-    return model.id;
-  },
-} satisfies SyncProvider<LLMGatewayModel>;
-
-// Every toggle reasoning control requires a leading wire-path comment, and the
-// sync runner only carries over headers that already exist on disk. Files this
-// sync writes with a toggle get the gateway-wide default; a hand-written
-// header on the existing file always wins.
-const TOGGLE_HEADER = `# Toggle: $.reasoning_effort = "none" disables thinking; any other accepted
-# value (or omitting the field) leaves it on. The gateway maps it to the
-# deployment's thinking switch.
-# https://docs.llmgateway.io/features/reasoning
-`;
-
-function toggleHeader(model: SyncedModel) {
-  return model.reasoning_options?.some((option) => option.type === "toggle")
-    ? TOGGLE_HEADER
-    : undefined;
-}
-
-// The LLM Gateway provider: one entry per upstream provider mapping, addressed
-// the way the gateway accepts provider-pinned requests (`provider/model-id`).
-export const llmgatewayProviders = {
-  id: "llmgateway-providers",
-  name: "LLM Gateway",
-  modelsDir: "providers/llmgateway-providers/models",
-  async fetchModels() {
-    return fetchLLMGatewayModels(`${API_ENDPOINT}?mapped=true`);
-  },
-  parseModels(raw) {
-    const data = LLMGatewayResponse.parse(raw).data;
-    // A deployment without the mapped view ignores the query param and returns
-    // aggregated root IDs (no provider prefix); syncing those here would wipe
-    // the provider-pinned catalog, so refuse to proceed. An empty response (or
-    // one left empty after filtering) would silently do the same via the
-    // delete-missing pass, so it is equally fatal.
-    if (data.length === 0 || !data.every((model) => model.id.includes("/"))) {
-      throw new Error("LLM Gateway mapped view unavailable: response is empty or contains unprefixed model ids");
-    }
-    // llmgateway/custom is the BYO-model placeholder and llmgateway/auto the
-    // auto-router; pinning either to a provider is meaningless in this catalog
-    // (the aggregated llmgateway provider carries `auto`).
-    const mapped = data.filter((model) => !model.id.startsWith("llmgateway/") && textOnly(model));
-    if (mapped.length === 0) {
-      throw new Error("LLM Gateway mapped view returned no text models");
-    }
-    // Every mapped entry is one specific provider deployment whose single
-    // providers[] mapping drives capabilities and reasoning controls. A kept
-    // entry with zero or several mappings would make the builder silently fall
-    // back to noisy supported_parameters / sibling defaults, so fail loudly.
-    const malformed = mapped.filter((model) => model.providers?.length !== 1);
-    if (malformed.length > 0) {
-      throw new Error(
-        `LLM Gateway mapped view returned entries without exactly one provider mapping: ${
-          malformed.map((model) => model.id).join(", ")
-        }`,
-      );
-    }
-    return mapped;
-  },
-  translateModel(model, context) {
-    const translated = buildLLMGatewayMappedModel(model, context.existing(model.id));
-    if (translated === undefined) {
-      return undefined;
-    }
-    return { id: model.id, model: translated, header: toggleHeader(translated) };
-  },
-  sourceID(model) {
-    return model.id;
+    return {
+      id: model.id,
+      model: buildLLMGatewayModel(model, context.existing(model.id)),
+    };
   },
 } satisfies SyncProvider<LLMGatewayModel>;
 
@@ -225,75 +125,12 @@ function modalities(values: string[], fallback: Modality[]): Modality[] {
   return [...new Set(result.length > 0 ? result : fallback)];
 }
 
-// Modalities as served by a specific deployment: a mapping without vision must
-// not carry image/pdf input, regardless of what the model-level architecture
-// claims — attachment=false with image input is contradictory.
-function deploymentModalities(model: LLMGatewayModel, vision: boolean | undefined) {
-  const base = defaultModalities(model);
-  if (vision !== false) {
-    return base;
-  }
-  const input = base.input.filter((value) => value !== "image" && value !== "pdf");
-  return {
-    input: input.length > 0 ? input : (["text"] satisfies Modality[]),
-    output: base.output,
-  };
-}
-
-const MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "models");
-const AGGREGATED_MODELS_DIR = path.join(import.meta.dirname, "..", "..", "..", "..", "..", "providers", "llmgateway", "models");
-const canonicalOutputLimitByID = new Map<string, number | undefined>();
-
-interface SiblingCuration {
-  reasoning_options?: SyncedFullModel["reasoning_options"];
-  interleaved?: SyncedFullModel["interleaved"];
-  cost_tiers?: NonNullable<SyncedFullModel["cost"]>["tiers"];
-}
-
-const siblingCurationByID = new Map<string, SiblingCuration>();
-
-// The aggregated llmgateway catalog curates reasoning controls, the reasoning
-// side-channel, and context pricing tiers for the same gateway surface; mapped
-// deployments of the same root model reuse them when the deployment does not
-// declare its own.
-function siblingCuration(rootID: string): SiblingCuration {
-  let curation = siblingCurationByID.get(rootID);
-  if (curation === undefined) {
-    const filePath = path.join(AGGREGATED_MODELS_DIR, `${rootID}.toml`);
-    const authored = existsSync(filePath)
-      ? Bun.TOML.parse(readFileSync(filePath, "utf8")) as SiblingCuration & {
-          cost?: { tiers?: NonNullable<SyncedFullModel["cost"]>["tiers"] };
-        }
-      : undefined;
-    curation = {
-      reasoning_options: authored?.reasoning_options?.length ? authored.reasoning_options : undefined,
-      interleaved: authored?.interleaved,
-      cost_tiers: authored?.cost?.tiers,
-    };
-    siblingCurationByID.set(rootID, curation);
-  }
-  return curation;
-}
-
-// Whether the canonical metadata declares limit.output; factored entries can
-// only omit their own output override when the base has one to inherit.
-function canonicalOutputLimit(modelID: string) {
-  if (!canonicalOutputLimitByID.has(modelID)) {
-    const filePath = path.join(MODELS_DIR, `${modelID}.toml`);
-    const metadata = existsSync(filePath)
-      ? Bun.TOML.parse(readFileSync(filePath, "utf8")) as { limit?: { output?: number } }
-      : undefined;
-    canonicalOutputLimitByID.set(modelID, metadata?.limit?.output);
-  }
-  return canonicalOutputLimitByID.get(modelID);
-}
-
-function resolveLLMGatewayBaseModel(model: LLMGatewayModel, modelID = model.id) {
-  const alias = BASE_MODEL_ALIASES[modelID];
+function resolveLLMGatewayBaseModel(model: LLMGatewayModel) {
+  const alias = BASE_MODEL_ALIASES[model.id];
   if (alias !== undefined) return alias;
   if (model.family === undefined) return undefined;
   const prefix = CANONICAL_FAMILY_ALIASES[model.family] ?? model.family;
-  return resolveModelMetadataBaseModel(`${prefix}/${modelID}`);
+  return resolveCanonicalBaseModel(`${prefix}/${model.id}`);
 }
 
 function inferFamily(model: LLMGatewayModel, name: string) {
@@ -315,19 +152,15 @@ function inferFamily(model: LLMGatewayModel, name: string) {
 export function buildLLMGatewayModel(
   model: LLMGatewayModel,
   existing: ExistingModel | undefined,
-): SyncedModel | undefined {
+): SyncedModel {
   const prompt = price(model.pricing.prompt);
   const completion = price(model.pricing.completion);
   const reasoning = model.supported_parameters.includes("reasoning")
     || model.supported_parameters.includes("include_reasoning");
   const reasoningOptions = llmGatewayReasoningOptions(model, existing);
-  const reported = model.context_length ?? 0;
-  // A missing/zero context must never be authored as limit.context = 0:
-  // factored entries leave it unset and inherit the base, and unfactored
-  // creates are skipped entirely. An authored 0 on the existing file is
-  // equally unusable and must not be re-stamped.
-  const servedContext = reported > 0 ? reported : undefined;
-  const context = servedContext ?? (existing?.limit?.context || undefined);
+  const context = model.context_length > 0
+    ? model.context_length
+    : existing?.limit?.context ?? model.context_length;
 
   // The gateway is authoritative for the volatile, gateway-specific data — cost,
   // served limits, and explicitly advertised reasoning efforts. Its
@@ -346,24 +179,15 @@ export function buildLLMGatewayModel(
         tiers: existing?.cost?.tiers,
       }
     : existing?.cost;
-  // Authored limits carry only known-positive values — never the zero/absent
-  // `reported` fallback.
-  const limit = context !== undefined
-    ? {
-        context,
-        input: existing?.limit?.input,
-        output: (existing?.limit?.output || undefined) ?? context,
-      }
-    : undefined;
+  const limit = {
+    context,
+    input: existing?.limit?.input,
+    output: existing?.limit?.output ?? context,
+  };
 
   // Existing factored model: refresh cost + limit, keep every authored override
   // as-is (undefined fields keep inheriting the base model).
   if (existing?.base_model !== undefined) {
-    const factoredLimit = {
-      context,
-      input: existing.limit?.input,
-      output: existing.limit?.output ?? context,
-    };
     return factorBaseModel(
       existing.base_model,
       {
@@ -376,7 +200,7 @@ export function buildLLMGatewayModel(
           tool_call: existing.tool_call,
           structured_output: existing.structured_output,
           open_weights: existing.open_weights,
-          limit: factoredLimit,
+          limit,
           modalities: existing.modalities,
         }),
         reasoning: existing.reasoning,
@@ -388,22 +212,16 @@ export function buildLLMGatewayModel(
         interleaved: existing.interleaved,
         knowledge: existing.knowledge,
         modalities: existing.modalities,
-        limit: factoredLimit,
+        limit,
         cost,
       },
-      factoredLimit,
+      limit,
       existing.base_model_omit,
     );
   }
 
   // Existing full model: refresh cost + limit, preserve curated metadata.
   if (existing !== undefined) {
-    // With no usable context from the API or the file there is nothing valid
-    // to author, and skipping would hand the file to the delete-missing pass —
-    // fail loudly rather than write limit.context = 0.
-    if (limit === undefined) {
-      throw new Error(`LLM Gateway entry ${model.id} has no usable context to author`);
-    }
     return {
       name: existing.name ?? model.name,
       description: existing.description ?? describeModel({
@@ -453,12 +271,7 @@ export function buildLLMGatewayModel(
   }
 
   // Brand-new model: best-effort translation from the gateway. Capability and
-  // modality data are unreliable here and should be hand-reviewed. Without a
-  // positive served context there is nothing usable to author, so skip.
-  if (servedContext === undefined) {
-    return undefined;
-  }
-  const createdLimit = limit ?? { context: servedContext, input: undefined, output: servedContext };
+  // modality data are unreliable here and should be hand-reviewed.
   const { input, output } = defaultModalities(model);
   return {
     name: model.name,
@@ -471,7 +284,7 @@ export function buildLLMGatewayModel(
         || model.supported_parameters.includes("tool_choice"),
       structured_output: model.structured_outputs ?? false,
       open_weights: false,
-      limit: createdLimit,
+      limit,
       modalities: { input, output },
     }),
     family: inferFamily(model, model.name),
@@ -486,261 +299,7 @@ export function buildLLMGatewayModel(
     structured_output: model.structured_outputs ?? false,
     open_weights: false,
     cost,
-    limit: createdLimit,
-    modalities: { input, output },
-  } satisfies SyncedFullModel;
-}
-
-export function buildLLMGatewayMappedModel(
-  model: LLMGatewayModel,
-  existing: ExistingModel | undefined,
-): SyncedModel | undefined {
-  // Mapped entries carry exactly one provider mapping; its capability flags
-  // describe that specific deployment, unlike the aggregated view where
-  // supported_parameters are too noisy to trust.
-  const mapping = model.providers?.[0];
-  const rootID = model.id.split("/").slice(1).join("/");
-  const prompt = price(model.pricing.prompt);
-  const completion = price(model.pricing.completion);
-  // The mapping's flag stays authoritative on resyncs too, so the written
-  // reasoning boolean and the reasoning_options derived from it always move
-  // together; prior curation only fills in when the mapping is silent, then
-  // the noisy supported_parameters signal as a last resort.
-  const reasoning = mapping?.reasoning
-    ?? existing?.reasoning
-    ?? (model.supported_parameters.includes("reasoning")
-      || model.supported_parameters.includes("include_reasoning"));
-  // The exact reasoning_effort values this deployment accepts. A deployment
-  // whose only accepted effort is "none" exposes a plain on/off switch (the
-  // gateway honours it through the thinking toggle), not effort tiers.
-  const deploymentOptions = mapping?.reasoning_efforts?.length
-    ? mapping.reasoning_efforts.length === 1 && mapping.reasoning_efforts[0] === "none"
-      ? [{ type: "toggle" as const }]
-      : [{ type: "effort" as const, values: mapping.reasoning_efforts }]
-    : undefined;
-  // Deployment-declared efforts own the effort/toggle surface; curation falls
-  // back from non-empty options on this file to the aggregated llmgateway
-  // catalog's controls for the same root model on the same gateway surface.
-  // Curated non-effort controls (e.g. budget_tokens for $.reasoning.max_tokens,
-  // which this host serves regardless of the effort list) survive alongside
-  // deployment efforts instead of being wiped by them. A curated [] counts as
-  // unknown so a bad first stamp is not sticky. Non-reasoning deployments
-  // carry none; the same applies to the interleaved reasoning side-channel.
-  const sibling = siblingCuration(rootID);
-  const curatedOptions = (existing?.reasoning_options?.length ? existing.reasoning_options : undefined)
-    ?? sibling.reasoning_options;
-  const reasoningOptions = reasoning
-    ? deploymentOptions !== undefined
-      ? [
-          ...(curatedOptions ?? []).filter((option) => option.type !== "effort" && option.type !== "toggle"),
-          ...deploymentOptions,
-        ]
-      : curatedOptions
-    : undefined;
-  const interleaved = reasoning
-    ? existing?.interleaved ?? sibling.interleaved
-    : undefined;
-  const reported = model.context_length ?? 0;
-  // Same zero-context rule as the aggregated builder: never author 0, inherit
-  // on factored entries, skip unfactored creates. An authored 0 on the
-  // existing file is equally unusable.
-  const servedContext = reported > 0 ? reported : undefined;
-  const context = servedContext ?? (existing?.limit?.context || undefined);
-
-  const cost = prompt !== undefined && completion !== undefined
-    ? {
-        input: prompt,
-        output: completion,
-        reasoning: reasoning ? nonZeroPrice(model.pricing.internal_reasoning) ?? existing?.cost?.reasoning : existing?.cost?.reasoning,
-        cache_read: nonZeroPrice(model.pricing.input_cache_read) ?? existing?.cost?.cache_read,
-        cache_write: nonZeroPrice(model.pricing.input_cache_write) ?? existing?.cost?.cache_write,
-        // The gateway API does not expose context pricing tiers, so authored
-        // tiers stick and new files seed from the aggregated sibling's curated
-        // tiers rather than silently under-stating long-context pricing.
-        tiers: existing?.cost?.tiers ?? sibling.cost_tiers,
-      }
-    : existing?.cost;
-  // The gateway's max_output is the deployment's real served limit, so it wins
-  // over inherited/authored values, unlike the aggregated view.
-  const servedOutput = (model.max_output || undefined) ?? (existing?.limit?.output || undefined);
-  // Authored limits carry only known-positive values — never the zero/absent
-  // `reported` fallback.
-  const limit = context !== undefined
-    ? {
-        context,
-        input: existing?.limit?.input,
-        output: servedOutput ?? context,
-      }
-    : undefined;
-
-  // Existing factored model: refresh cost + limit, keep every authored override
-  // as-is. Unlike the aggregated provider, the name override must be carried
-  // forward: mapped names disambiguate deployments of the same model (e.g.
-  // "GPT-5.5 (Azure)" vs "GPT-5.5 (OpenAI)") and must not collapse back to the
-  // base metadata name.
-  if (existing?.base_model !== undefined) {
-    // Mirror the brand-new factored path: without a served or authored output,
-    // keep inheriting the base's output rather than stamping context over it.
-    const factoredLimit = {
-      context,
-      input: existing.limit?.input,
-      output: servedOutput ?? (canonicalOutputLimit(existing.base_model) !== undefined ? undefined : context),
-    };
-    // Deployment capability flags keep their create-path authority on
-    // resyncs: a mapping that gains or loses reasoning/vision/tools/structured
-    // outputs realigns the written flags together with the reasoning_options
-    // computed from them, instead of freezing stale curation forever.
-    return factorBaseModel(
-      existing.base_model,
-      {
-        name: existing.name ?? model.name,
-        attachment: mapping?.vision ?? existing.attachment,
-        // No describeModel fallback: synthesizing a description here would
-        // stamp a sticky generic override on every name-pinned factored entry;
-        // leaving it unset keeps inheriting the lab text from the base.
-        description: existing.description,
-        reasoning: mapping?.reasoning ?? existing.reasoning,
-        reasoning_options: reasoningOptions,
-        temperature: existing.temperature,
-        tool_call: mapping?.tools ?? existing.tool_call,
-        structured_output: model.structured_outputs ?? existing.structured_output,
-        status: existing.status,
-        interleaved,
-        knowledge: existing.knowledge,
-        // Vision realigns modalities in both directions: false strips
-        // image/pdf, true clears any stale stripped override so the base's
-        // richer inputs inherit again; only a silent mapping keeps curation.
-        modalities: mapping?.vision === undefined
-          ? existing.modalities
-          : mapping.vision
-            ? undefined
-            : deploymentModalities(model, false),
-        limit: factoredLimit,
-        cost,
-      },
-      factoredLimit,
-      existing.base_model_omit,
-    );
-  }
-
-  // Existing full model: refresh cost + limit, preserve curated metadata.
-  // Capability flags follow the same rule as the factored path above: the
-  // deployment mapping wins, curation fills the gaps.
-  if (existing !== undefined) {
-    // With no usable context from the API or the file there is nothing valid
-    // to author, and skipping would hand the file to the delete-missing pass —
-    // fail loudly rather than write limit.context = 0.
-    if (limit === undefined) {
-      throw new Error(`LLM Gateway mapped entry ${model.id} has no usable context to author`);
-    }
-    const resolved = {
-      attachment: mapping?.vision ?? existing.attachment ?? false,
-      tool_call: mapping?.tools ?? existing.tool_call ?? false,
-      structured_output: model.structured_outputs ?? existing.structured_output,
-      // Same bidirectional vision rule as the factored path; with no base to
-      // inherit from, a declared vision recomputes from the served
-      // architecture instead of clearing.
-      modalities: mapping?.vision === undefined
-        ? existing.modalities ?? deploymentModalities(model, undefined)
-        : deploymentModalities(model, mapping.vision),
-    };
-    return {
-      name: existing.name ?? model.name,
-      description: existing.description ?? describeModel({
-        id: model.id,
-        name: existing.name ?? model.name,
-        family: existing.family,
-        reasoning,
-        tool_call: resolved.tool_call,
-        structured_output: resolved.structured_output,
-        open_weights: existing.open_weights,
-        limit,
-        modalities: resolved.modalities,
-      }),
-      family: existing.family,
-      release_date: existing.release_date ?? dateFromTimestamp(model.created),
-      last_updated: existing.last_updated ?? dateFromTimestamp(model.created),
-      attachment: resolved.attachment,
-      reasoning,
-      reasoning_options: reasoningOptions,
-      temperature: existing.temperature ?? false,
-      tool_call: resolved.tool_call,
-      structured_output: resolved.structured_output,
-      knowledge: existing.knowledge,
-      open_weights: existing.open_weights ?? false,
-      status: existing.status,
-      interleaved,
-      cost,
-      limit,
-      modalities: resolved.modalities,
-    } satisfies SyncedFullModel;
-  }
-
-  // Brand-new model with a reviewed metadata entry: factor against the
-  // canonical base. The mapped ID is `serving-provider/model-id` and the
-  // serving provider is unrelated to the originating lab, so resolve the base
-  // from the root model ID + family, and keep the disambiguating name. The
-  // mapping's own capability flags describe this specific deployment, so they
-  // go in as overrides (factorBaseModel drops the ones equal to the base).
-  const canonical = resolveLLMGatewayBaseModel(model, rootID);
-  if (canonical !== undefined) {
-    const factoredLimit = {
-      context,
-      input: undefined,
-      // Without a served limit, inherit the base's output; only fall back to
-      // context when the base declares none (output is required downstream).
-      output: model.max_output ?? (canonicalOutputLimit(canonical) !== undefined ? undefined : context),
-    };
-    return factorBaseModel(canonical, {
-      name: model.name,
-      attachment: mapping?.vision,
-      reasoning: mapping?.reasoning,
-      reasoning_options: reasoningOptions,
-      interleaved,
-      tool_call: mapping?.tools,
-      structured_output: model.structured_outputs,
-      // A deployment without vision must not inherit image/pdf inputs from
-      // the base — attachment=false with image input is contradictory.
-      modalities: mapping?.vision === false ? deploymentModalities(model, false) : undefined,
-      limit: factoredLimit,
-      cost,
-    }, factoredLimit);
-  }
-
-  // Brand-new model without metadata: best-effort translation. The mapping's
-  // own capability flags are reliable here; modalities mirror the mapping too.
-  // Without a positive served context there is nothing usable to author.
-  if (servedContext === undefined) {
-    return undefined;
-  }
-  const createdLimit = limit ?? { context: servedContext, input: undefined, output: servedOutput ?? servedContext };
-  const { input, output } = deploymentModalities(model, mapping?.vision);
-  return {
-    name: model.name,
-    description: describeModel({
-      id: model.id,
-      name: model.name,
-      family: inferFamily(model, model.name),
-      reasoning,
-      tool_call: mapping?.tools ?? false,
-      structured_output: model.structured_outputs ?? false,
-      open_weights: false,
-      limit: createdLimit,
-      modalities: { input, output },
-    }),
-    family: inferFamily(model, model.name),
-    release_date: dateFromTimestamp(model.created),
-    last_updated: dateFromTimestamp(model.created),
-    attachment: mapping?.vision ?? input.some((value) => value !== "text"),
-    reasoning,
-    reasoning_options: reasoningOptions,
-    interleaved,
-    temperature: model.supported_parameters.includes("temperature"),
-    tool_call: mapping?.tools ?? false,
-    structured_output: model.structured_outputs ?? false,
-    open_weights: false,
-    cost,
-    limit: createdLimit,
+    limit,
     modalities: { input, output },
   } satisfies SyncedFullModel;
 }
@@ -749,7 +308,7 @@ function llmGatewayReasoningOptions(
   model: LLMGatewayModel,
   existing: ExistingModel | undefined,
 ): SyncedFullModel["reasoning_options"] {
-  const advertised = new Set((model.providers ?? []).flatMap((provider) => provider.reasoning_efforts ?? []));
+  const advertised = new Set(model.providers.flatMap((provider) => provider.reasoning_efforts ?? []));
   if (advertised.size === 0) return undefined;
 
   const efforts = [...advertised].sort((a, b) => {

packages/core/src/sync/providers/openrouter.ts

@@ -68,7 +68,7 @@ export const OpenRouterModel = z.object({
     input_cache_read: z.string().optional(),
     input_cache_write: z.string().optional(),
     overrides: z.array(z.object({
-      min_prompt_tokens: z.number().optional(),
+      min_prompt_tokens: z.number(),
       prompt: z.string().optional(),
       completion: z.string().optional(),
       input_cache_read: z.string().optional(),
@@ -159,7 +159,7 @@ function costTiers(model: OpenRouterModel, existing: ExistingModel | undefined)
     .flatMap((o) => {
       const input = price(o.prompt);
       const output = price(o.completion);
-      if (o.min_prompt_tokens === undefined || input === undefined || output === undefined) return [];
+      if (input === undefined || output === undefined) return [];
       return [{
         tier: { type: "context" as const, size: o.min_prompt_tokens },
         input,

packages/core/src/sync/providers/vercel.ts

@@ -69,17 +69,9 @@ export const vercel = {
     return VercelResponse.parse(raw).data;
   },
   translateModel(model, context) {
-    const existing = context.existing(model.id);
-    const routeBase = freeRouteBase(model.id);
-    const baseModel = existing?.base_model ?? resolveVercelBaseModel(model.id);
-    const inherited = routeBase === undefined ? undefined : context.existing(routeBase);
     return {
       id: model.id,
-      model: buildVercelModel(
-        model,
-        existing,
-        inherited ?? (baseModel === undefined || baseModel === model.id ? undefined : context.existing(baseModel)),
-      ),
+      model: buildVercelModel(model, context.existing(model.id)),
     };
   },
   sameModel(current, desired) {
@@ -87,11 +79,7 @@ export const vercel = {
   },
 } satisfies SyncProvider<VercelModel>;
 
-export function buildVercelModel(
-  model: VercelModel,
-  existing: ExistingModel | undefined,
-  base: ExistingModel | undefined = undefined,
-): SyncedModel {
+export function buildVercelModel(model: VercelModel, existing: ExistingModel | undefined): SyncedModel {
   const tags = new Set(model.tags);
   const releaseDate = model.released
     ? dateFromTimestamp(model.released)
@@ -145,7 +133,7 @@ export function buildVercelModel(
     last_updated: existing?.last_updated ?? releaseDate,
     attachment: existing?.attachment ?? (tags.has("vision") || tags.has("file-input")),
     reasoning: existing?.reasoning ?? tags.has("reasoning"),
-    reasoning_options: existing?.reasoning_options ?? base?.reasoning_options,
+    reasoning_options: existing?.reasoning_options,
     temperature: existing?.temperature,
     tool_call: model.type === "language"
       ? existing?.tool_call ?? tags.has("tool-use")
@@ -180,35 +168,11 @@ export function buildVercelModel(
     },
   };
 
-  const baseModel = existing?.base_model ?? resolveVercelBaseModel(model.id);
+  const baseModel = existing?.base_model ?? resolveCanonicalBaseModel(model.id);
   if (baseModel === undefined) return synced;
 
-  return factorBaseModel(baseModel, {
-    name: synced.name,
-    attachment: synced.attachment,
-    reasoning: synced.reasoning,
-    reasoning_options: synced.reasoning_options,
-    temperature: synced.temperature,
-    tool_call: synced.tool_call,
-    structured_output: synced.structured_output,
-    status: synced.status,
-    interleaved: synced.interleaved,
-    experimental: synced.experimental,
-    provider: synced.provider,
-    cost: synced.cost,
-    limit: synced.limit,
-    modalities: synced.modalities,
-  }, synced.limit, existing?.base_model_omit);
-}
-
-function resolveVercelBaseModel(modelID: string) {
-  const routeBase = freeRouteBase(modelID);
-  return resolveCanonicalBaseModel(modelID)
-    ?? (routeBase === undefined ? undefined : resolveCanonicalBaseModel(routeBase));
-}
-
-function freeRouteBase(modelID: string) {
-  return modelID.endsWith("-free") ? modelID.slice(0, -"-free".length) : undefined;
+  const { last_updated: _lastUpdated, ...overrides } = synced;
+  return factorBaseModel(baseModel, overrides, synced.limit, existing?.base_model_omit);
 }
 
 function dateFromTimestamp(timestamp: number) {

packages/core/test/auto-merge.test.ts

@@ -82,7 +82,7 @@ test("does not inspect deleted models", async () => {
 });
 
 test("allows reviewed providers with explicit reasoning options", async () => {
-  for (const provider of ["crossmodel", "edenai", "empiriolabs", "hyper", "kilo", "llmgateway", "llmgateway-providers", "merge-gateway", "nano-gpt", "openrouter", "venice"]) {
+  for (const provider of ["empiriolabs", "kilo", "llmgateway", "merge-gateway", "nano-gpt", "openrouter", "venice"]) {
     const decision = await classifyAutoMerge(
       [{ status: "updated", path: `providers/${provider}/models/reasoner.toml` }],
       async () => fullModel(true, 'reasoning_options = [{ type = "toggle" }]'),

packages/core/test/baseten.test.ts

@@ -1,48 +0,0 @@
-import { expect, test } from "bun:test";
-
-import {
-  buildBasetenModel,
-  type BasetenModel,
-} from "../src/sync/providers/baseten.js";
-
-function basetenModel(overrides: Partial<BasetenModel> = {}): BasetenModel {
-  return {
-    id: "deepseek-ai/DeepSeek-V4-Flash-0731",
-    name: "DeepSeek V4 Flash 0731",
-    context_length: 1_048_576,
-    max_completion_tokens: 1_048_576,
-    input_modalities: ["text"],
-    output_modalities: ["text"],
-    pricing: { prompt: "0.00000013", completion: "0.00000026" },
-    supported_features: ["reasoning", "tools", "structured_outputs"],
-    supported_sampling_parameters: ["temperature"],
-    ...overrides,
-  };
-}
-
-test("preserves an explicitly authored Baseten output limit", () => {
-  const built = buildBasetenModel(
-    basetenModel(),
-    undefined,
-    "deepseek/deepseek-v4-flash-0731",
-    { limit: { context: 1_048_576, output: 384_000 } },
-  );
-
-  expect(built).toMatchObject({
-    base_model: "deepseek/deepseek-v4-flash-0731",
-    limit: { context: 1_048_576, output: 384_000 },
-  });
-});
-
-test("uses Baseten's catalog output limit without an authored override", () => {
-  const built = buildBasetenModel(
-    basetenModel({ max_completion_tokens: 262_144 }),
-    undefined,
-    "deepseek/deepseek-v4-pro-0813",
-  );
-
-  expect(built).toMatchObject({
-    base_model: "deepseek/deepseek-v4-pro-0813",
-    limit: { context: 1_048_576, output: 262_144 },
-  });
-});

packages/core/test/sync.test.ts

@@ -10,7 +10,7 @@ import {
   parseAnthropicPricing,
   type AnthropicModel,
 } from "../src/sync/providers/anthropic.js";
-import { buildCortecsModel, cortecs, type CortecsModel } from "../src/sync/providers/cortecs.js";
+import { buildCortecsModel, type CortecsModel } from "../src/sync/providers/cortecs.js";
 import {
   buildCrossModel,
   CrossModelResponse,
@@ -56,13 +56,7 @@ import {
   resolveCanonicalBaseModel,
   type OpenRouterModel,
 } from "../src/sync/providers/openrouter.js";
-import {
-  buildLLMGatewayMappedModel,
-  buildLLMGatewayModel,
-  llmgateway,
-  llmgatewayProviders,
-  type LLMGatewayModel,
-} from "../src/sync/providers/llmgateway.js";
+import { buildLLMGatewayModel, type LLMGatewayModel } from "../src/sync/providers/llmgateway.js";
 import {
   buildMergeGatewayModel,
   fetchMergeGatewayModels,
@@ -473,46 +467,6 @@ test("parses CrossModel's nullable reasoning controls", () => {
   });
 });
 
-test("syncs CrossModel's explicit reasoning controls", () => {
-  const model = buildCrossModel(
-    crossModelModel({
-      capabilities: {
-        json: true,
-        reasoning: {
-          supported: true,
-          toggle: true,
-          effort: ["low", "high", "max"],
-          budget_tokens: { min: 1_024, max: 32_000 },
-        },
-      },
-    }),
-    undefined,
-  );
-
-  expect(model).toMatchObject({
-    reasoning_options: [
-      { type: "toggle" },
-      { type: "effort", values: ["low", "high", "max"] },
-      { type: "budget_tokens", min: 1_024, max: 32_000 },
-    ],
-  });
-});
-
-test("rejects unknown CrossModel reasoning efforts", () => {
-  expect(() =>
-    CrossModelResponse.parse({
-      data: [
-        {
-          ...crossModelModel(),
-          capabilities: {
-            reasoning: { supported: true, effort: ["unexpected"] },
-          },
-        },
-      ],
-    })
-  ).toThrow();
-});
-
 test("syncs NanoGPT's verified reasoning, pricing, limits, and open-weight metadata", () => {
   const model = buildNanoGptModel(nanoGptModel({
     pricing: {
@@ -2080,33 +2034,6 @@ test("OpenRouter sync maps pricing.overrides into cost tiers", () => {
   });
 });
 
-test("OpenRouter sync ignores time-window pricing overrides", () => {
-  const source = openRouterModel({
-    pricing: {
-      prompt: "0.00000132",
-      completion: "0.00000396",
-      overrides: [{
-        utc_start: 1_000,
-        utc_end: 100,
-        prompt: "0.00000066",
-        completion: "0.00000198",
-      }],
-    },
-  });
-  const [parsed] = openrouter.parseModels({ data: [source] });
-  const model = buildOpenRouterModel(parsed!, {
-    cost: {
-      input: 1.32,
-      output: 3.96,
-      tiers: [{ tier: { type: "context", size: 200_000 }, input: 2.64, output: 7.92 }],
-    },
-  });
-
-  expect(model.cost?.tiers).toEqual([
-    { tier: { type: "context", size: 200_000 }, input: 2.64, output: 7.92 },
-  ]);
-});
-
 test("OpenRouter sync keeps authored tiers when API omits overrides", () => {
   const model = buildOpenRouterModel(openRouterModel({
     pricing: {
@@ -2681,27 +2608,6 @@ test("defaults new reasoning models to empty reasoning options", () => {
   });
 });
 
-test("inherits base reasoning options instead of stamping empty ones", () => {
-  expect(preserveReasoningOptions({ reasoning: true }, undefined, undefined, [{ type: "toggle" }]))
-    .toEqual({ reasoning: true });
-});
-
-test("normalizes Cortecs file modalities to pdf", () => {
-  const [model] = cortecs.parseModels({
-    object: "list",
-    data: [{
-      id: "document-model",
-      created: 1_775_088_000,
-      pricing: { currency: "EUR", input_token: 1, output_token: 2 },
-      context_size: 65_536,
-      input_modalities: ["text", "file"],
-      output_modalities: ["text"],
-    }],
-  });
-
-  expect(model.input_modalities).toEqual(["text", "pdf"]);
-});
-
 test("preserves authored Cortecs reasoning options missing from the API", () => {
   const model: CortecsModel = {
     id: "deepseek-v4-flash-0731",
@@ -3032,52 +2938,6 @@ test("factors aliased LLM Gateway routes against canonical metadata", () => {
   });
 });
 
-test("factors mapped LLM Gateway entries against the root model metadata", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel(), undefined);
-
-  expect(model).toEqual({
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
-    structured_output: true,
-    cost: {
-      input: 10,
-      output: 50,
-      cache_read: 1,
-      cache_write: 12.5,
-    },
-  });
-});
-
-test("applies deployment capability flags on mapped factored entries", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    providers: [{ providerId: "anthropic", vision: false, tools: false, reasoning: false }],
-    architecture: { input_modalities: ["text"], output_modalities: ["text"] },
-    max_output: 64_000,
-  }), undefined);
-
-  expect(model).toEqual({
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    attachment: false,
-    reasoning: false,
-    tool_call: false,
-    structured_output: true,
-    modalities: {
-      input: ["text"],
-    },
-    limit: {
-      output: 64_000,
-    },
-    cost: {
-      input: 10,
-      output: 50,
-      cache_read: 1,
-      cache_write: 12.5,
-    },
-  });
-});
-
 test("factors Grok LLM Gateway routes against xAI metadata", () => {
   const model = buildLLMGatewayModel(llmGatewayModel({
     id: "grok-4-6",
@@ -3101,339 +2961,6 @@ test("factors Grok LLM Gateway routes against xAI metadata", () => {
   });
 });
 
-test("prefers the gateway max_output over authored output on mapped resyncs", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({ max_output: 32_000 }), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    description: "Claude Fable 5 served by Anthropic",
-    limit: { output: 64_000 },
-  });
-
-  expect(model).toEqual({
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    description: "Claude Fable 5 served by Anthropic",
-    reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
-    structured_output: true,
-    limit: {
-      output: 32_000,
-    },
-    cost: {
-      input: 10,
-      output: 50,
-      cache_read: 1,
-      cache_write: 12.5,
-    },
-  });
-});
-
-test("translates a none-only effort list into a reasoning toggle", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    providers: [{ providerId: "anthropic", vision: true, tools: true, reasoning: true, reasoning_efforts: ["none"] }],
-  }), undefined);
-
-  expect(model).toMatchObject({
-    base_model: "anthropic/claude-fable-5",
-    reasoning_options: [{ type: "toggle" }],
-  });
-});
-
-test("realigns capability flags from the mapping on mapped factored resyncs", () => {
-  // The deployment dropped reasoning and gained tools since the file was
-  // written: the resync must move the booleans and the reasoning controls
-  // together instead of clearing options under a frozen reasoning = true.
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    providers: [{ providerId: "anthropic", vision: true, tools: true, reasoning: false }],
-  }), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    reasoning: true,
-    reasoning_options: [{ type: "toggle" }],
-    tool_call: false,
-  });
-
-  expect(model).toMatchObject({ reasoning: false });
-  expect(model!.reasoning_options).toBeUndefined();
-  // Realigned to the mapping and now equal to the base, the stale
-  // tool_call = false override is dropped and inherits the base again.
-  expect(model!.tool_call).toBeUndefined();
-});
-
-test("restores image input when vision returns on mapped resyncs", () => {
-  // The file was written while the deployment had no vision (text-only
-  // stripped modalities); vision is back, so the stale override must clear.
-  const factored = buildLLMGatewayMappedModel(llmGatewayMappedModel(), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    attachment: false,
-    modalities: { input: ["text"] },
-  });
-  expect(factored!.modalities).toBeUndefined();
-  expect(factored!.attachment).toBeUndefined();
-
-  const full = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "acme/mystery-model",
-    name: "Mystery Model (Acme)",
-    family: undefined,
-    providers: [{ providerId: "acme", vision: true, tools: true, reasoning: false }],
-  }), {
-    name: "Mystery Model (Acme)",
-    attachment: false,
-    modalities: { input: ["text"], output: ["text"] },
-  });
-  expect(full).toMatchObject({
-    attachment: true,
-    modalities: { input: ["text", "image"], output: ["text"] },
-  });
-});
-
-test("never synthesizes a description on mapped factored resyncs", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel(), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-  });
-
-  // An unset description must keep inheriting the base's lab text instead of
-  // being stamped with a sticky synthesized override on the first resync.
-  expect(model).toBeDefined();
-  expect(model!.description).toBeUndefined();
-});
-
-test("authors the toggle wire-path header on mapped sync creates", () => {
-  const context = { existing: () => undefined, authored: () => undefined };
-
-  const toggle = llmgatewayProviders.translateModel(llmGatewayMappedModel({
-    providers: [{ providerId: "anthropic", vision: true, tools: true, reasoning: true, reasoning_efforts: ["none"] }],
-  }), context);
-  expect(toggle?.header).toStartWith("# Toggle: $.reasoning_effort");
-
-  const effort = llmgatewayProviders.translateModel(llmGatewayMappedModel(), context);
-  expect(effort?.model).toMatchObject({
-    reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
-  });
-  expect(effort?.header).toBeUndefined();
-});
-
-test("keeps inheriting base output on factored resyncs without max_output", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({ max_output: undefined }), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    description: "Claude Fable 5 served by Anthropic",
-  });
-
-  expect(model).toEqual({
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    description: "Claude Fable 5 served by Anthropic",
-    reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
-    structured_output: true,
-    cost: {
-      input: 10,
-      output: 50,
-      cache_read: 1,
-      cache_write: 12.5,
-    },
-  });
-});
-
-test("skips unfactorable LLM Gateway creates without a served context", () => {
-  // Unknown family, so no canonical base to inherit a context from.
-  const mapped = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "acme/mystery-model",
-    name: "Mystery Model (Acme)",
-    family: undefined,
-    context_length: undefined,
-  }), undefined);
-  expect(mapped).toBeUndefined();
-
-  const aggregated = buildLLMGatewayModel(llmGatewayModel({
-    id: "mystery-model",
-    name: "Mystery Model",
-    family: undefined,
-    context_length: undefined,
-  }), undefined);
-  expect(aggregated).toBeUndefined();
-});
-
-test("keeps curated budget controls under deployment efforts", () => {
-  // Deployment efforts own only the effort/toggle surface: the hand-authored
-  // budget_tokens control (this host's $.reasoning.max_tokens path) survives
-  // the resync, while the stale effort list is replaced.
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel(), {
-    base_model: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    reasoning_options: [
-      { type: "effort", values: ["low", "high"] },
-      { type: "budget_tokens", min: 1_024, max: 63_999 },
-    ],
-  });
-  expect(model!.reasoning_options).toEqual([
-    { type: "budget_tokens", min: 1_024, max: 63_999 },
-    { type: "effort", values: ["low", "medium", "high", "xhigh", "max"] },
-  ]);
-
-  // Same merge on creates, with the budget coming from the aggregated
-  // sibling's curation for the same root model.
-  const seeded = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "anthropic/claude-sonnet-4-6",
-    name: "Claude Sonnet 4.6 (Anthropic)",
-  }), undefined);
-  expect(seeded!.reasoning_options).toEqual([
-    { type: "budget_tokens", min: 1_024, max: 63_999 },
-    { type: "effort", values: ["low", "medium", "high", "xhigh", "max"] },
-  ]);
-});
-
-test("seeds context pricing tiers from the aggregated sibling on creates", () => {
-  // The gateway API carries no tier pricing; without the sibling's curated
-  // tiers the bulk sync would author tiered models at flat long-context rates.
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "openai/gpt-5.5",
-    name: "GPT-5.5 (OpenAI)",
-    family: "openai",
-  }), undefined);
-
-  expect(model!.cost?.tiers).toEqual([
-    { tier: { type: "context", size: 272_000 }, input: 10, output: 45, cache_read: 1 },
-  ]);
-});
-
-test("factors perplexity entries without widening the shared prefix map", () => {
-  // The perplexity family resolves through resolveModelMetadataBaseModel's
-  // exact models/ path match; CANONICAL_PROVIDER_PREFIXES stays untouched so
-  // other hosts' standalone perplexity files keep their current behavior.
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "perplexity/sonar-pro",
-    name: "Sonar Pro (Perplexity)",
-    family: "perplexity",
-  }), undefined);
-
-  expect(model).toMatchObject({ base_model: "perplexity/sonar-pro" });
-});
-
-test("refuses to author a zero context on full LLM Gateway resyncs", () => {
-  // Existing full rows (no base to inherit from) with nothing usable from the
-  // API or the file must fail loudly instead of being rewritten with
-  // limit.context = 0.
-  expect(() => buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    context_length: undefined,
-    max_output: undefined,
-  }), {
-    name: "Claude Fable 5 (Anthropic)",
-  })).toThrow("no usable context");
-
-  // An authored 0 on disk is as unusable as an absent context.
-  expect(() => buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    context_length: 0,
-    max_output: undefined,
-  }), {
-    name: "Claude Fable 5 (Anthropic)",
-    limit: { context: 0 },
-  })).toThrow("no usable context");
-
-  expect(() => buildLLMGatewayModel(llmGatewayModel({
-    context_length: undefined,
-  }), {
-    name: "Claude Fable 5",
-  })).toThrow("no usable context");
-});
-
-test("leaves context unset on mapped factored creates without a served context", () => {
-  const model = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    context_length: undefined,
-    max_output: undefined,
-  }), undefined);
-
-  // Everything limit-related inherits from the base; no zero is authored.
-  expect(model).toBeDefined();
-  expect("limit" in model!).toBe(false);
-});
-
-test("strips image input when the deployment has no vision", () => {
-  // The model-level architecture still claims image input; the deployment
-  // flag must win on both the factored and the unfactored path.
-  const factored = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    providers: [{ providerId: "anthropic", vision: false, tools: true, reasoning: false }],
-  }), undefined);
-  expect(factored).toMatchObject({
-    base_model: "anthropic/claude-fable-5",
-    attachment: false,
-    modalities: { input: ["text"] },
-  });
-
-  const full = buildLLMGatewayMappedModel(llmGatewayMappedModel({
-    id: "acme/mystery-model",
-    name: "Mystery Model (Acme)",
-    family: undefined,
-    providers: [{ providerId: "acme", vision: false, tools: true, reasoning: false }],
-  }), undefined);
-  expect(full).toMatchObject({
-    attachment: false,
-    modalities: { input: ["text"], output: ["text"] },
-  });
-});
-
-test("refuses empty responses in both LLM Gateway syncs", () => {
-  expect(() => llmgateway.parseModels({ data: [] })).toThrow("no text models");
-  expect(() => llmgatewayProviders.parseModels({ data: [] })).toThrow("mapped view unavailable");
-});
-
-test("refuses aggregated responses in the mapped LLM Gateway sync", () => {
-  expect(() => llmgatewayProviders.parseModels({ data: [llmGatewayModel()] }))
-    .toThrow("mapped view unavailable");
-});
-
-test("filters pseudo and non-text entries from the mapped LLM Gateway sync", () => {
-  const parsed = llmgatewayProviders.parseModels({
-    data: [
-      llmGatewayMappedModel(),
-      llmGatewayMappedModel({ id: "llmgateway/auto", name: "Auto Route (LLM Gateway)" }),
-      llmGatewayMappedModel({
-        id: "openai/sora-2",
-        name: "Sora 2 (OpenAI)",
-        architecture: { input_modalities: ["text"], output_modalities: ["video"] },
-      }),
-    ],
-  });
-
-  expect(parsed.map((model) => model.id)).toEqual(["anthropic/claude-fable-5"]);
-});
-
-test("refuses mapped LLM Gateway entries without exactly one provider mapping", () => {
-  expect(() => llmgatewayProviders.parseModels({
-    data: [llmGatewayMappedModel({ providers: undefined })],
-  })).toThrow("without exactly one provider mapping");
-
-  expect(() => llmgatewayProviders.parseModels({
-    data: [llmGatewayMappedModel({ providers: [] })],
-  })).toThrow("without exactly one provider mapping");
-
-  expect(() => llmgatewayProviders.parseModels({
-    data: [
-      llmGatewayMappedModel(),
-      llmGatewayMappedModel({
-        id: "azure/gpt-5.5",
-        name: "GPT-5.5 (Azure)",
-        providers: [{ providerId: "azure" }, { providerId: "openai" }],
-      }),
-    ],
-  })).toThrow("azure/gpt-5.5");
-
-  // Entries the sync drops anyway (pseudo-models, non-text) may lack a
-  // mapping without tripping the guard.
-  const parsed = llmgatewayProviders.parseModels({
-    data: [
-      llmGatewayMappedModel(),
-      llmGatewayMappedModel({
-        id: "llmgateway/auto",
-        name: "Auto Route (LLM Gateway)",
-        providers: undefined,
-      }),
-    ],
-  });
-  expect(parsed.map((model) => model.id)).toEqual(["anthropic/claude-fable-5"]);
-});
-
 // Ensures catalog pagination preserves authentication and returns every page.
 test("fetches every page of the Merge Gateway catalog", async () => {
   const requests: string[] = [];
@@ -3987,44 +3514,6 @@ test("Vercel factored models inherit temperature from base metadata", () => {
   expect(synced).not.toHaveProperty("temperature");
 });
 
-test("Vercel free routes factor onto the canonical non-free model", () => {
-  const [model] = vercel.parseModels({
-    data: [{
-      id: "zai/glm-4.6v-flash-free",
-      name: "GLM-4.6V-Flash (Free)",
-      created: 1_765_152_000,
-      released: 1_765_152_000,
-      context_window: 128_000,
-      max_tokens: 24_000,
-      type: "language",
-      tags: ["reasoning", "tool-use", "vision", "file-input"],
-      pricing: { input: "0", output: "0" },
-    }],
-  });
-
-  const translated = vercel.translateModel(model!, {
-    existing(id) {
-      return id === "zai/glm-4.6v-flash"
-        ? { reasoning_options: [{ type: "toggle" }] }
-        : undefined;
-    },
-    authored() {
-      return undefined;
-    },
-  });
-
-  expect(translated?.model).toMatchObject({
-    base_model: "zhipuai/glm-4.6v-flash",
-    name: "GLM-4.6V-Flash (Free)",
-    reasoning_options: [{ type: "toggle" }],
-    cost: { input: 0, output: 0 },
-    limit: { output: 24_000 },
-    modalities: { input: ["text", "image", "pdf"] },
-  });
-  expect(translated?.model).not.toHaveProperty("description");
-  expect(translated?.model).not.toHaveProperty("family");
-});
-
 test("Vercel Claude Opus fast variants factor onto base opus metadata", () => {
   const [model] = vercel.parseModels({
     data: [{
@@ -4044,26 +3533,11 @@ test("Vercel Claude Opus fast variants factor onto base opus metadata", () => {
     }],
   });
 
-  const translated = vercel.translateModel(model!, {
-    existing(id) {
-      return id === "anthropic/claude-opus-5"
-        ? { reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }] }
-        : undefined;
-    },
-    authored() {
-      return undefined;
-    },
-  });
-  const synced = translated?.model;
-
-  expect(synced).toMatchObject({
+  expect(buildVercelModel(model!, undefined)).toMatchObject({
     base_model: "anthropic/claude-opus-5",
     name: "Claude Opus 5 (Fast)",
-    reasoning_options: [{ type: "effort", values: ["low", "medium", "high", "xhigh", "max"] }],
     cost: { input: 10, output: 50, cache_read: 1, cache_write: 12.5 },
   });
-  expect(synced).not.toHaveProperty("description");
-  expect(synced).not.toHaveProperty("family");
 });
 
 test("OpenRouter Claude Opus fast variants factor onto base opus metadata", () => {
@@ -4303,22 +3777,6 @@ function llmGatewayModel(overrides: Partial<LLMGatewayModel> = {}): LLMGatewayMo
   };
 }
 
-function llmGatewayMappedModel(overrides: Partial<LLMGatewayModel> = {}): LLMGatewayModel {
-  return llmGatewayModel({
-    id: "anthropic/claude-fable-5",
-    name: "Claude Fable 5 (Anthropic)",
-    providers: [{
-      providerId: "anthropic",
-      vision: true,
-      tools: true,
-      reasoning: true,
-      reasoning_efforts: ["low", "medium", "high", "xhigh", "max"],
-    }],
-    max_output: 128_000,
-    ...overrides,
-  });
-}
-
 function mergeGatewayVendor(
   overrides: Partial<MergeGatewayModel["vendors"][string]> = {},
 ): MergeGatewayModel["vendors"][string] {

packages/sdk/src/types.ts

@@ -68,10 +68,7 @@ export interface CostTier extends Cost {
 
 /** Pricing for a provider's model, including context-size tiers. */
 export interface ModelCost extends Cost {
-  /**
-   * Legacy compatibility field for context-tier pricing.
-   * @deprecated Use `tiers` to read the exact context threshold.
-   */
+  /** Legacy compatibility field: pricing applied beyond 200K context. Prefer `tiers`. */
   context_over_200k?: Cost
   /** Context-size-based pricing tiers. */
   tiers?: CostTier[]

providers/agentrouter/logo.svg

@@ -1,4 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor" role="img">
-  <title>AgentRouter</title>
-  <path d="M11.8 5C7.8 4.2 3 9.6 4.2 15.6c.4 3 3 4.6 6.6 3-1.8-3-2.6-7.8 1-13.6Zm.4 14c4 .8 8.8-4.6 7.6-10.6-.4-3-3-4.6-6.6-3 1.8 3 2.6 7.8-1 13.6ZM12 9.5c.2 1.5 1 2.3 2.5 2.5-1.5.2-2.3 1-2.5 2.5-.2-1.5-1-2.3-2.5-2.5 1.5-.2 2.3-1 2.5-2.5Z"/>
-</svg>

providers/agentrouter/models/claude-opus-4-8.toml

@@ -1,8 +0,0 @@
-# Catalog and endpoint support: https://agentrouter.org/api/pricing
-# AgentRouter publishes relative token ratios, not USD prices, so cost is intentionally omitted.
-base_model = "anthropic/claude-opus-4-8"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-
-[provider]
-npm = "@ai-sdk/anthropic"
-api = "https://agentrouter.org/v1"

providers/agentrouter/models/claude-opus-5.toml

@@ -1,8 +0,0 @@
-# Catalog and endpoint support: https://agentrouter.org/api/pricing
-# AgentRouter publishes relative token ratios, not USD prices, so cost is intentionally omitted.
-base_model = "anthropic/claude-opus-5"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-
-[provider]
-npm = "@ai-sdk/anthropic"
-api = "https://agentrouter.org/v1"

providers/agentrouter/models/gpt-5.6-sol.toml

@@ -1,4 +0,0 @@
-# Catalog and endpoint support: https://agentrouter.org/api/pricing
-# AgentRouter publishes relative token ratios, not USD prices, so cost is intentionally omitted.
-base_model = "openai/gpt-5.6-sol"
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]

providers/agentrouter/provider.toml

@@ -1,5 +0,0 @@
-name = "AgentRouter"
-env = ["AGENTROUTER_API_KEY"]
-npm = "@ai-sdk/openai-compatible"
-api = "https://agentrouter.org/v1"
-doc = "https://agentrouter.org/docs/opencode.html"

providers/aki-io/models/deepseek-v4-flash-0731-284b.toml

@@ -1,13 +0,0 @@
-# attachment are not supported on AKI.IO for security reasons and therefore are disabled
-base_model = "deepseek/deepseek-v4-flash-0731"
-reasoning_options = [{ type = "effort", values = ["none", "low", "high", "max"] }]
-
-# cost as described on the AKI.IO pricing page at https://aki.io/#pricing
-[cost]
-input = 0.20
-output = 0.50
-cache_read = 0.10
-
-[limit]
-context = 1_048_576
-output = 81_920

providers/aki-io/models/gemma4-26b.toml

@@ -2,6 +2,7 @@
 # AKI.IO uses the reasoning_option 'none' to disable reasoning
 base_model = "google/gemma-4-26b-a4b-it"
 attachment = false
+reasoning = true
 reasoning_options = [{ type = "effort", values = ["none", "high"] }]
 
 [cost]

providers/aki-io/models/kimi-k2.7-code-1100b.toml

@@ -4,11 +4,9 @@ base_model = "moonshotai/kimi-k2.7-code"
 attachment = false
 reasoning_options = []
 
-# cost as described on the AKI.IO pricing page at https://aki.io/#pricing
 [cost]
 input = 0.86
 output = 3.00
-cache_read = 0.18
 
 [limit]
 context = 262_144

providers/aki-io/models/minimax-m2.5-230b.toml

@@ -0,0 +1,12 @@
+# attachment are not supported on AKI.IO for security reasons and therefore are disabled
+base_model = "minimax/MiniMax-M2.5"
+attachment = false 
+reasoning_options = []
+
+[cost]
+input = 0.25
+output = 1.20
+
+[limit]
+context = 196_608
+output = 65_536

providers/aki-io/models/qwen3.8-27b.toml

@@ -1,14 +0,0 @@
-# attachment are not supported on AKI.IO for security reasons and therefore are disabled
-# AKI.IO uses the reasoning_option 'none' to disable reasoning
-base_model = "alibaba/qwen3.8-27b"
-attachment = false
-reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "xhigh"] }]
-
-# cost as described on the AKI.IO pricing page at https://aki.io/#pricing
-[cost]
-input = 0.30
-output = 2.20
-cache_read = 0.10
-
-[modalities]
-input = ["text", "image"]

providers/alibaba-cn/models/qwen-plus.toml

@@ -21,8 +21,6 @@ max = 81_920
 input = 0.115
 output = 0.287
 reasoning = 1.147
-cache_read = 0.012
-cache_write = 0.144
 
 [limit]
 context = 1_000_000

providers/amazon-bedrock/models/global.openai.gpt-5.6-luna.toml

@@ -1,20 +1,19 @@
-# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
-# Verified from eu-west-1: returns 200 with reasoning effort set.
-# Pricing mirrors the US in-region on-demand rates on openai.gpt-5.6-luna: https://aws.amazon.com/bedrock/pricing/
+# bedrock-runtime global CRIS ID (not Mantle). Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
+# Global CRIS pricing (USD per 1M tokens): short ≤272K / long >272K — input / 30m cache write / cache read / output
 base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-5.6 Luna (Global)"
 
 [cost]
-input = 0.22
-output = 1.32
-cache_read = 0.022
-cache_write = 0.275
+input = 0.20
+output = 1.20
+cache_read = 0.02
+cache_write = 0.25
 
 [[cost.tiers]]
 tier = { size = 272_000 }
-input = 0.44
-output = 1.98
-cache_read = 0.044
-cache_write = 0.55
+input = 0.40
+output = 1.80
+cache_read = 0.04
+cache_write = 0.50

providers/amazon-bedrock/models/global.openai.gpt-5.6-sol.toml

@@ -1,20 +1,19 @@
-# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
-# Verified from eu-west-1: returns 200 with reasoning effort set.
-# Pricing mirrors the US in-region on-demand rates on openai.gpt-5.6-sol: https://aws.amazon.com/bedrock/pricing/
+# bedrock-runtime global CRIS ID (not Mantle). Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
+# Global CRIS pricing (USD per 1M tokens): short ≤272K / long >272K — input / 30m cache write / cache read / output
 base_model = "openai/gpt-5.6-sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-5.6 Sol (Global)"
 
 [cost]
-input = 5.50
-output = 33.00
-cache_read = 0.55
-cache_write = 6.875
+input = 5.00
+output = 30.00
+cache_read = 0.50
+cache_write = 6.25
 
 [[cost.tiers]]
 tier = { size = 272_000 }
-input = 11.00
-output = 49.50
-cache_read = 1.10
-cache_write = 13.75
+input = 10.00
+output = 45.00
+cache_read = 1.00
+cache_write = 12.50

providers/amazon-bedrock/models/global.openai.gpt-5.6-terra.toml

@@ -1,20 +1,19 @@
-# Global cross-Region inference profile — served by Bedrock core (Converse), not Mantle, so no [provider] override.
-# Verified from eu-west-1: effort none|low|medium|high|xhigh|max all return 200.
-# Pricing mirrors the US in-region on-demand rates on openai.gpt-5.6-terra: https://aws.amazon.com/bedrock/pricing/
+# bedrock-runtime global CRIS ID (not Mantle). Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
+# Global CRIS pricing (USD per 1M tokens): short ≤272K / long >272K — input / 30m cache write / cache read / output
 base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
 base_model_omit = ["cost.context_over_200k", "experimental.modes.fast"]
 name = "GPT-5.6 Terra (Global)"
 
 [cost]
-input = 2.20
-output = 13.20
-cache_read = 0.22
-cache_write = 2.75
+input = 2.00
+output = 12.00
+cache_read = 0.20
+cache_write = 2.50
 
 [[cost.tiers]]
 tier = { size = 272_000 }
-input = 4.40
-output = 19.80
-cache_read = 0.44
-cache_write = 5.50
+input = 4.00
+output = 18.00
+cache_read = 0.40
+cache_write = 5.00

providers/amazon-bedrock/models/openai.gpt-5.4.toml

@@ -1,3 +1,5 @@
+# Mantle in-region only in us-east-1/us-east-2/us-west-2 (plus gov); pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-54.html
 base_model = "openai/gpt-5.4"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 base_model_omit = ["cost.tiers", "cost.context_over_200k", "experimental.modes.fast", "limit.input"]
@@ -14,5 +16,5 @@ output = 128_000
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/openai.gpt-5.5.toml

@@ -1,3 +1,5 @@
+# Mantle in-region only in us-east-1/us-east-2; pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-55.html
 base_model = "openai/gpt-5.5"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh"] }]
 base_model_omit = ["cost.tiers", "cost.context_over_200k", "experimental.modes.fast", "limit.input"]
@@ -14,5 +16,5 @@ output = 128_000
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/openai.gpt-5.6-luna.toml

@@ -1,6 +1,6 @@
+# Mantle in-region only in us-east-1/us-east-2/us-west-2; pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card (regions, mantle URL, in-region pricing): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-luna.html
 # Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
-# Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing (US East N. Virginia & Ohio; also US West Oregon): https://aws.amazon.com/bedrock/pricing/
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-luna"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
@@ -21,5 +21,5 @@ cache_write = 0.55
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/openai.gpt-5.6-sol.toml

@@ -1,6 +1,6 @@
+# Mantle in-region only in us-east-1/us-east-2; pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card (regions, mantle URL, in-region pricing): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-sol.html
 # Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
-# Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing (US East N. Virginia & Ohio): https://aws.amazon.com/bedrock/pricing/
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-sol"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
@@ -21,5 +21,5 @@ cache_write = 13.75
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/openai.gpt-5.6-terra.toml

@@ -1,6 +1,6 @@
+# Mantle in-region only in us-east-1/us-east-2/us-west-2; pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card (regions, mantle URL, in-region pricing): https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-56-terra.html
 # Bedrock model IDs and reasoning support: https://github.com/openai/codex/pull/30285
-# Bedrock context window: https://developers.openai.com/api/docs/guides/amazon-bedrock
-# In-region on-demand pricing (US East N. Virginia & Ohio; also US West Oregon): https://aws.amazon.com/bedrock/pricing/
 # Columns: input / 30m cache write / cache read / output (USD per 1M tokens)
 base_model = "openai/gpt-5.6-terra"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high", "xhigh", "max"] }]
@@ -21,5 +21,5 @@ cache_write = 5.50
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/xai.grok-4.3.toml

@@ -1,3 +1,5 @@
+# Mantle in-region only in us-east-1/us-east-2/us-west-2; pin us-east-1 so non-US AWS_REGION does not 404.
+# Model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-3.html
 base_model = "xai/grok-4.3"
 reasoning_options = [{ type = "effort", values = ["none", "low", "medium", "high"] }]
 last_updated = "2026-06-28"
@@ -16,5 +18,5 @@ input = ["text", "image"]
 
 [provider]
 npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
+api = "https://bedrock-mantle.us-east-1.api.aws/openai/v1"
 shape = "responses"

providers/amazon-bedrock/models/xai.grok-4.6.toml

@@ -1,14 +0,0 @@
-# Source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-xai-grok-4-6.html
-base_model = "xai/grok-4.6"
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh"] }]
-last_updated = "2026-08-18"
-
-[cost]
-input = 2.20
-output = 6.60
-cache_read = 0.55
-
-[provider]
-npm = "@ai-sdk/amazon-bedrock/mantle"
-api = "https://bedrock-mantle.${AWS_REGION}.api.aws/openai/v1"
-shape = "responses"

providers/ambient/models/ambient/large.toml

@@ -7,9 +7,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 0.6
-output = 2
-cache_read = 0.15
+input = 1.05
+output = 4.4
+cache_read = 0.2
 cache_write = 0
 
 [limit]

providers/ambient/models/deepseek/deepseek-v4-flash-0731.toml

@@ -11,9 +11,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 0.08
-output = 0.18
-cache_read = 0.016
+input = 0.14
+output = 0.28
+cache_read = 0.028
 cache_write = 0
 
 [limit]

providers/ambient/models/z-ai/glm-5.2.toml

@@ -6,9 +6,9 @@ type = "effort"
 values = ["high", "max"]
 
 [cost]
-input = 0.6
-output = 2
-cache_read = 0.15
+input = 1.05
+output = 4.4
+cache_read = 0.2
 cache_write = 0
 
 [limit]

providers/amd/logo.svg

@@ -1,3 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor">
-  <path d="M18.324 9.137l1.559 1.56h2.556v2.557L24 14.814V9.137zM2 9.52l-2 4.96h1.309l.37-.982H3.9l.408.982h1.338L3.432 9.52zm4.209 0v4.955h1.238v-3.092l1.338 1.562h.188l1.338-1.556v3.091h1.238V9.52H10.47l-1.592 1.845L7.287 9.52zm6.283 0v4.96h2.057c1.979 0 2.88-1.046 2.88-2.472 0-1.36-.937-2.488-2.747-2.488zm1.237.91h.792c1.17 0 1.63.711 1.63 1.57 0 .728-.372 1.572-1.616 1.572h-.806zm-10.985.273l.791 1.932H2.008zm17.137.307l-1.604 1.603v2.25h2.246l1.604-1.607h-2.246z"/>
-</svg>

providers/amd/models/DeepSeek-V4-Flash.toml

@@ -1,16 +0,0 @@
-base_model = "deepseek/deepseek-v4-flash-0731"
-
-# Reasoning control verified against the AMD Token Factory API (accessed 2026-08-16):
-# POST https://developer.amd.com.cn/radeon/api/v1/chat/completions validates
-# reasoning_effort with enum none|minimal|low|medium|high|xhigh|max (invalid
-# values return 400). thinking.* and budget controls are silently ignored and no
-# reasoning_content is ever returned, so there is no toggle and no interleaved
-# channel on this host.
-reasoning_options = [
-  { type = "effort", values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"] },
-]
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.0028

providers/amd/provider.toml

@@ -1,6 +0,0 @@
-name = "AMD"
-env = ["AMD_API_KEY"]
-npm = "@ai-sdk/openai-compatible"
-# Token Factory (BETA) / Radeon Cloud, China endpoint
-api = "https://developer.amd.com.cn/radeon/api/v1"
-doc = "https://developer.amd.com.cn/radeon/tokenfactory"

providers/arcee/logo.svg

@@ -1,3 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="currentColor" fill-rule="evenodd">
-  <path d="M13.236 2.377 2.751 20.493H0L11.863 0l1.373 2.377Zm3.554 6.156-9.606 11.96H4.13L15.511 6.32l1.279 2.212Zm6.908 11.96H14.05l8.406-2.151 1.242 2.15Zm-3.42-5.922-7.843 5.92H8.482l10.597-7.997 1.2 2.077Z"/>
-</svg>

providers/arcee/models/deepseek/deepseek-v4-flash-latest.toml

@@ -1,19 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cache hit: $0.028/MTok (Arcee correction, 2026-08-14).
-# No toggle: thinking.type=enabled|disabled is rejected by the live API.
-# Effort: reasoning_effort = high|max; low is rejected.
-base_model = "deepseek/deepseek-v4-flash-0731"
-name = "DeepSeek V4 Flash Latest"
-status = "beta"
-reasoning_options = [{ type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.14
-output = 0.28
-cache_read = 0.028
-
-[limit]
-context = 1_048_576

providers/arcee/models/deepseek/deepseek-v4-pro-0813.toml

@@ -1,18 +0,0 @@
-# Live inventory and pricing: https://api.arcee.ai/api/v1/models (verified 2026-08-17).
-# This dated route is not yet listed on Arcee's public pricing page.
-# No toggle: thinking.type=enabled|disabled is rejected by the live API.
-# Effort: reasoning_effort = high|max; low is rejected.
-base_model = "deepseek/deepseek-v4-pro-0813"
-status = "beta"
-reasoning_options = [{ type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.32
-output = 3.96
-cache_read = 0.044
-
-[limit]
-context = 1_048_576

providers/arcee/models/deepseek/deepseek-v4-pro.toml

@@ -1,18 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cached input pricing: https://docs.arcee.ai/get-started/pricing.md
-# No toggle: thinking.type=enabled|disabled is rejected by the live API.
-# Effort: reasoning_effort = high|max
-base_model = "deepseek/deepseek-v4-pro"
-status = "beta"
-reasoning_options = [{ type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.74
-output = 3.48
-cache_read = 0.20
-
-[limit]
-context = 512_000

providers/arcee/models/moonshotai/kimi-k3.toml

@@ -1,19 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cached input pricing: https://docs.arcee.ai/get-started/pricing.md
-# Effort: reasoning_effort = low|high|max
-base_model = "moonshotai/kimi-k3"
-attachment = false
-temperature = true
-status = "beta"
-reasoning_options = [{ type = "effort", values = ["low", "high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 3.00
-output = 15.00
-cache_read = 0.30
-
-[modalities]
-input = ["text"]

providers/arcee/models/thinkingmachines/inkling-small.toml

@@ -1,22 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cached input pricing: https://docs.arcee.ai/get-started/pricing.md
-# A live reasoning_effort request is rejected; reasoning is always on.
-base_model = "thinkingmachines/inkling-small"
-attachment = false
-status = "beta"
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.50
-output = 1.20
-cache_read = 0.10
-
-[limit]
-context = 262_144
-output = 262_144
-
-[modalities]
-input = ["text"]

providers/arcee/models/trinity-large-thinking.toml

@@ -1,17 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cached input pricing: https://docs.arcee.ai/get-started/pricing.md
-# A live reasoning_effort request is rejected; reasoning is always on.
-base_model = "arcee-ai/trinity-large-thinking"
-status = "beta"
-reasoning_options = []
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 0.25
-output = 0.80
-cache_read = 0.06
-
-[limit]
-context = 262_144

providers/arcee/models/zai-org/glm-5.2.toml

@@ -1,17 +0,0 @@
-# Beta catalog and pricing: https://www.arcee.ai/blog/open-models-api-beta
-# Cached input pricing: https://docs.arcee.ai/get-started/pricing.md
-# Effort: reasoning_effort = high|max
-base_model = "zhipuai/glm-5.2"
-status = "beta"
-reasoning_options = [{ type = "effort", values = ["high", "max"] }]
-
-[interleaved]
-field = "reasoning_content"
-
-[cost]
-input = 1.40
-output = 4.40
-cache_read = 0.26
-
-[limit]
-context = 262_144

providers/arcee/provider.toml

@@ -1,5 +0,0 @@
-name = "Arcee"
-npm = "@ai-sdk/openai-compatible"
-env = ["ARCEE_API_KEY"]
-api = "https://api.arcee.ai/api/v1"
-doc = "https://docs.arcee.ai"

providers/azure-cognitive-services/models/claude-opus-4-7.toml

@@ -1,18 +0,0 @@
-# Azure AI Foundry serves Claude Opus 4.7 at Anthropic list pricing:
-# https://platform.claude.com/docs/en/about-claude/pricing
-# Claude 4.6+ bills the full 1M context window at standard rates (no long-context
-# premium), so no [[cost.tiers]] here. Model availability:
-# https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/claude-models?tabs=pay-go#available-claude-models
-base_model = "anthropic/claude-opus-4-7"
-# Possible values described at https://platform.claude.com/docs/en/build-with-claude/effort#effort-levels
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 5.00
-output = 25.00
-cache_read = 0.50
-cache_write = 6.25
-
-[provider]
-npm = "@ai-sdk/anthropic"
-api = "https://${AZURE_COGNITIVE_SERVICES_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/azure/models/claude-opus-4-7.toml

@@ -1,19 +0,0 @@
-# Azure AI Foundry serves Claude Opus 4.7 at Anthropic list pricing:
-# https://platform.claude.com/docs/en/about-claude/pricing
-# Claude 4.6+ bills the full 1M context window at standard rates (no long-context
-# premium), so no [[cost.tiers]] here — the tiers on azure's claude-opus-4-6/4-8
-# entries predate that policy. Model availability:
-# https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/claude-models?tabs=pay-go#available-claude-models
-base_model = "anthropic/claude-opus-4-7"
-# Possible values described at https://platform.claude.com/docs/en/build-with-claude/effort#effort-levels
-reasoning_options = [{ type = "effort", values = ["low", "medium", "high", "xhigh", "max"] }]
-
-[cost]
-input = 5.00
-output = 25.00
-cache_read = 0.50
-cache_write = 6.25
-
-[provider]
-npm = "@ai-sdk/anthropic"
-api = "https://${AZURE_RESOURCE_NAME}.services.ai.azure.com/anthropic/v1"

providers/baseten/models/deepseek-ai/DeepSeek-V4-Flash-0731.toml

@@ -1,19 +1,16 @@
-# Baseten documents top-level reasoning_effort = none | minimal | low | medium |
-# high (default) | xhigh | max for DeepSeek V4 Flash 0731. Reasoning output is
-# returned in reasoning_content.
-# Baseten serves at most 384k output tokens for this model; keep this explicit
-# because its catalog has reported the 1,048,576-token context as the output cap.
+# Baseten's model matrix documents DeepSeek V4 Flash 0731 reasoning as "Enabled
+# by default", but the model is absent from the "Control reasoning depth" table
+# on the reasoning page, and that page warns that models outside it accept
+# reasoning_effort and ignore it. So it reasons with no addressable depth.
 # https://docs.baseten.co/inference/model-apis/overview
 # https://docs.baseten.co/inference/model-apis/reasoning
 base_model = "deepseek/deepseek-v4-flash-0731"
+name = "Deepseek V4 Flash 0731"
+reasoning_options = []
 
 [interleaved]
 field = "reasoning_content"
 
-[[reasoning_options]]
-type = "effort"
-values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
-
 [cost]
 input = 0.13
 output = 0.26
@@ -21,4 +18,4 @@ cache_read = 0.028
 
 [limit]
 context = 1_048_576
-output = 384_000
+output = 1_048_576

providers/baseten/models/deepseek-ai/DeepSeek-V4-Pro-0813.toml

@@ -1,15 +1,13 @@
-# Baseten documents top-level reasoning_effort = none | low | high | max for
-# DeepSeek V4 Pro 0813. Reasoning output is returned in reasoning_content.
+# Baseten documents top-level reasoning_effort = none | minimal | low |
+# medium (default) | high | xhigh | max for DeepSeek V4 Pro.
 # https://docs.baseten.co/inference/model-apis/reasoning
 base_model = "deepseek/deepseek-v4-pro-0813"
+name = "Deepseek V4 Pro 0813"
 description = "Flagship DeepSeek model for coding, reasoning, and agentic work"
 
-[interleaved]
-field = "reasoning_content"
-
 [[reasoning_options]]
 type = "effort"
-values = ["none", "low", "high", "max"]
+values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 [cost]
 input = 1.32

providers/baseten/models/deepseek-ai/DeepSeek-V4-Pro.toml

@@ -2,7 +2,9 @@
 # medium (default) | high | xhigh | max for DeepSeek V4 Pro; no toggle or
 # reasoning-token budget field is documented.
 # https://docs.baseten.co/inference/model-apis/reasoning
+
 base_model = "deepseek/deepseek-v4-pro"
+name = "Deepseek V4 Pro"
 
 [interleaved]
 field = "reasoning_content"
@@ -17,5 +19,5 @@ output = 3.48
 cache_read = 0.145
 
 [limit]
-context = 1_048_576
+context = 262_144
 output = 262_144

providers/baseten/models/moonshotai/Kimi-K2.7-Code.toml

@@ -1,14 +1,9 @@
 # Baseten documents opt-in via chat_template_args.enable_thinking=true, but no
-# explicit false behavior, effort values, or reasoning-token budget. Reasoning
-# output is returned in reasoning_content.
+# explicit false behavior, effort values, or reasoning-token budget.
 # https://docs.baseten.co/inference/model-apis/reasoning
 
 base_model = "moonshotai/kimi-k2.7-code"
 temperature = true
-
-[interleaved]
-field = "reasoning_content"
-
 [[reasoning_options]]
 type = "toggle"
 

providers/baseten/models/moonshotai/Kimi-K3.toml

@@ -1,15 +1,12 @@
 # Baseten documents Kimi K3 reasoning as enabled by default, with top-level
 # reasoning_effort = none | low | high | max (default). It also accepts
 # chat_template_args.enable_thinking=false. No reasoning-token budget is
-# documented. Reasoning output is returned in reasoning_content.
+# documented.
 # https://docs.baseten.co/inference/model-apis/reasoning
 base_model = "moonshotai/kimi-k3"
 description = "Kimi multimodal agent model for visual understanding, coding, and planning"
 temperature = true
 
-[interleaved]
-field = "reasoning_content"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "low", "high", "max"]

providers/baseten/models/openai/gpt-oss-120b.toml

@@ -1,7 +1,6 @@
 # Baseten now documents top-level reasoning_effort = none | minimal | low |
 # medium (default) | high | xhigh | max for GPT OSS 120B; reasoning is
-# otherwise enabled by default. No toggle or budget exists. Reasoning output is
-# returned in reasoning_content.
+# otherwise enabled by default. No toggle or budget exists.
 # https://docs.baseten.co/inference/model-apis/reasoning
 
 name = "OpenAI GPT 120B"
@@ -17,9 +16,6 @@ structured_output = true
 knowledge = "2025-08"
 open_weights = true
 
-[interleaved]
-field = "reasoning_content"
-
 [[reasoning_options]]
 type = "effort"
 values = ["none", "minimal", "low", "medium", "high", "xhigh", "max"]

providers/baseten/models/zai-org/GLM-5.2-Fast.toml

@@ -7,7 +7,6 @@
 # https://docs.baseten.co/inference/model-apis/reasoning
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2 Fast"
-attachment = true
 
 [interleaved]
 field = "reasoning_content"
@@ -24,6 +23,3 @@ cache_read = 0.21
 [limit]
 context = 1_048_576
 output = 262_144
-
-[modalities]
-input = ["text", "image"]

providers/baseten/models/zai-org/GLM-5.2.toml

@@ -5,7 +5,6 @@
 # https://docs.baseten.co/inference/model-apis/reasoning
 base_model = "zhipuai/glm-5.2"
 name = "GLM 5.2"
-attachment = true
 
 [interleaved]
 field = "reasoning_content"
@@ -22,6 +21,3 @@ cache_read = 0.3
 [limit]
 context = 1_048_576
 output = 262_144
-
-[modalities]
-input = ["text", "image"]

providers/cerebras/models/zai-glm-4.7.toml

@@ -0,0 +1,33 @@
+name = "Z.AI GLM-4.7"
+description = "Flagship GLM model for hybrid reasoning, coding, and agentic engineering"
+# Model-specific reasoning HTTP values (accessed 2026-06-25):
+# Reasoning defaults on; reasoning_effort = "none" disables it. The alternative
+# disable_reasoning = true|false is deprecated and will be removed after
+# 2026-07-21. No positive effort levels or token budget are documented.
+# Sources:
+# https://inference-docs.cerebras.ai/capabilities/reasoning#glm-reasoning-effort-and-disable-reasoning
+# https://inference-docs.cerebras.ai/models/zai-glm-47
+release_date = "2026-01-07"
+last_updated = "2026-06-10"
+attachment = false
+reasoning = true
+reasoning_options = [{ type = "effort", values = ["none"] }]
+temperature = true
+tool_call = true
+open_weights = true
+structured_output = true
+status = "beta"
+
+[cost]
+input = 2.25
+output = 2.75
+cache_read = 2.25
+cache_write = 0
+
+[limit]
+context = 131_072
+output = 40_960
+
+[modalities]
+input = ["text"]
+output = ["text"]

providers/chutes/models/Qwen/Qwen3.8-27B-TEE.toml

@@ -1,17 +0,0 @@
-base_model = "alibaba/qwen3.8-27b"
-name = "Qwen3.8 27B TEE"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-release_date = "2026-08-16"
-last_updated = "2026-08-16"
-reasoning_options = []
-
-[cost]
-input = 0.35
-output = 2.75
-cache_read = 0.03499999999999999
-
-[limit]
-output = 65_536
-
-[modalities]
-input = ["text", "image"]

providers/cloudflare-ai-gateway/models/openai/gpt-4o-mini.toml

@@ -1,6 +1,6 @@
 base_model = "openai/gpt-4o-mini"
 
 [cost]
-input = 0.075
-output = 0.3
-cache_read = 0.0375
+input = 0.15
+output = 0.6
+cache_read = 0.075

providers/cloudflare-ai-gateway/models/openai/gpt-4o.toml

@@ -1,6 +1,6 @@
 base_model = "openai/gpt-4o"
 
 [cost]
-input = 1.25
-output = 5
-cache_read = 0.625
+input = 2.5
+output = 10
+cache_read = 1.25

providers/cloudflare-workers-ai/models/@cf/deepseek-ai/deepseek-v4-flash-0731.toml

@@ -2,12 +2,7 @@
 # Effort: reasoning_effort = high|max.
 base_model = "deepseek/deepseek-v4-flash-0731"
 
-[[reasoning_options]]
-type = "toggle"
-
-[[reasoning_options]]
-type = "effort"
-values = ["high", "max"]
+reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["high", "max"] }]
 
 [cost]
 input = 0.44
@@ -15,5 +10,5 @@ output = 1.32
 cache_read = 0.014
 
 [limit]
-context = 1_310_720
+context = 1_048_576
 output = 1_048_576

providers/cloudflare-workers-ai/models/@cf/qwen/qwen3.8-27b.toml

@@ -1,19 +0,0 @@
-# Workers AI exposes `reasoning_effort` and `chat_template_kwargs`; Qwen supports
-# enable_thinking plus low|medium|xhigh effort levels.
-# https://developers.cloudflare.com/workers-ai/models/qwen3.8-27b/ (accessed 2026-08-18)
-# https://huggingface.co/Qwen/Qwen3.8-27B (accessed 2026-08-18)
-
-base_model = "alibaba/qwen3.8-27b"
-description = "Qwen vision-language model for visual reasoning, documents, and agent tasks"
-reasoning_options = [{ type = "toggle" }, { type = "effort", values = ["low", "medium", "xhigh"] }]
-
-[cost]
-input = 0.45
-output = 3.2
-cache_read = 0.05
-
-[limit]
-output = 262_144
-
-[modalities]
-input = ["text", "image"]

providers/cortecs/models/claude-4-5-sonnet.toml

@@ -18,3 +18,6 @@ cache_write = 4.078
 
 [limit]
 output = 200_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-4-6-sonnet.toml

@@ -18,3 +18,6 @@ cache_write = 3.999
 
 [limit]
 output = 1_000_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-haiku-4-5.toml

@@ -18,3 +18,6 @@ cache_write = 1.186
 
 [limit]
 output = 200_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-opus-5.toml

@@ -17,3 +17,6 @@ cache_write = 6.874
 
 [limit]
 output = 1_000_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-opus4-5.toml

@@ -18,3 +18,6 @@ cache_write = 6.645
 
 [limit]
 output = 200_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-opus4-6.toml

@@ -18,3 +18,6 @@ cache_write = 6.645
 
 [limit]
 output = 1_000_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-opus4-7.toml

@@ -18,3 +18,6 @@ cache_write = 6.797
 
 [limit]
 output = 1_000_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-opus4-8.toml

@@ -17,3 +17,6 @@ cache_write = 6.797
 
 [limit]
 output = 1_000_000
+
+[modalities]
+input = ["text", "image"]

providers/cortecs/models/claude-sonnet-4.toml

@@ -18,3 +18,6 @@ cache_write = 3.624
 
 [limit]
 output = 200_000
+
+[modalities]
+input = ["text", "image"]