Update Groq models: Add Qwen 3.6 27b and ALLaM 2 7b (#3761)

* Update Groq models

* fix(groq): add missing cost block to allam-2-7b

* fix(groq): refine ALLaM 2 7b pricing source comment

* fix(groq): verify ALLaM 2 7b free tier pricing

* fix(groq): align ALLaM comment placement and pricing link

models/sdaia/allam-2-7b.toml

@@ -0,0 +1,21 @@
+name = "ALLaM-2-7b"
+description = "ALLaM-2-7b instruction tuned model by SDAIA"
+release_date = "2025-01-23"
+last_updated = "2025-01-23"
+attachment = false
+reasoning = false
+temperature = true
+tool_call = false
+open_weights = true
+
+[limit]
+context = 4096
+output = 4096
+
+[modalities]
+input = ["text"]
+output = ["text"]
+
+[[weights]]
+label = "Hugging Face"
+url = "https://huggingface.co/ALLaM-AI/ALLaM-2.0-7B-Instruct"

providers/groq/models/allam-2-7b.toml

@@ -0,0 +1,7 @@
+# ALLaM-2-7B is provided unbilled via the Groq developer tier.
+# Verified via Groq Models API (no pricing object) and https://groq.com/pricing
+base_model = "sdaia/allam-2-7b"
+
+[cost]
+input = 0
+output = 0

providers/groq/models/meta-llama/llama-4-scout-17b-16e-instruct.toml

@@ -1,25 +0,0 @@
-name = "Llama 4 Scout 17B 16E"
-description = "Open multimodal Llama model for long-context analysis and efficient agents"
-family = "llama"
-release_date = "2025-04-05"
-last_updated = "2025-04-05"
-attachment = true
-reasoning = false
-temperature = true
-knowledge = "2024-08"
-tool_call = true
-structured_output = true
-open_weights = true
-status = "beta"
-
-[cost]
-input = 0.11
-output = 0.34
-
-[limit]
-context = 131_072
-output = 8_192
-
-[modalities]
-input = ["text", "image"]
-output = ["text"]

providers/groq/models/qwen/qwen3-32b.toml

@@ -1,32 +0,0 @@
-# Reasoning HTTP format (accessed 2026-06-25):
-# POST https://api.groq.com/openai/v1/chat/completions
-# The model page recommends reasoning_effort: "default" for thinking and calls
-# the model dual-mode, but the API page scopes "none"/"default" to Qwen 3.6 27B.
-# Raw HTTP acceptance of both values for Qwen3-32B remains unresolved.
-# Sources:
-# https://console.groq.com/docs/model/qwen/qwen3-32b
-# https://console.groq.com/docs/reasoning
-name = "Qwen3-32B"
-description = "Qwen instruction model for multilingual chat, reasoning, and tool use"
-family = "qwen"
-release_date = "2025-06-11"
-last_updated = "2025-06-12"
-attachment = false
-reasoning = true
-reasoning_options = [{ type = "effort", values = ["none", "default"] }]
-temperature = true
-tool_call = true
-open_weights = true
-status = "beta"
-
-[cost]
-input = 0.29
-output = 0.59
-
-[limit]
-context = 131_072
-output = 40_960
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/groq/models/qwen/qwen3.6-27b.toml

@@ -0,0 +1,15 @@
+base_model = "alibaba/qwen3.6-27b"
+reasoning_options = [{ type = "effort", values = ["none", "default"] }]
+
+[cost]
+input = 0.60
+output = 3.00
+cache_read = 0.30
+
+[limit]
+context = 131_072
+output = 16_384
+
+[modalities]
+input = ["text", "image"]
+output = ["text"]