chore(vultr): sync model catalog to latest API

- Update Kimi-K2.6 pricing to 0.30/1.20 (was 0.15/0.60)
- Replace GLM-5.1-FP8 with GLM-5.2-FP8 (base_model zhipuai/glm-5.2,
  limit.context 393_216 per vLLM max_model_len)
- Add Qwen3.5-397B-A17B (0.30/2.00)
- Add Qwen3.6-27B (0.30/2.00)
- Add MiMo-V2.5-Pro (0.55/1.65)
- Add DeepSeek-V4-Flash (0.30/1.00)
- Remove Llama-3.1-Nemotron-Safety-Guard-8B-v3 (no longer in API)

All reasoning models declare reasoning_options = [] per Vultr's
lack of reasoning request control. New TextGeneration models use
base_model inheritance from models/ metadata. ReRank models
(VultronRetriever*) are intentionally skipped pending official
documentation of their capabilities.

Source: https://api.vultrinference.com/v1/models
Co-authored-by: Cursor <cursoragent@cursor.com>

providers/vultr/models/Qwen/Qwen3.5-397B-A17B.toml

@@ -0,0 +1,6 @@
+base_model = "alibaba/qwen3.5-397b-a17b"
+reasoning_options = []
+
+[cost]
+input = 0.30
+output = 2.00

providers/vultr/models/Qwen/Qwen3.6-27B.toml

@@ -0,0 +1,6 @@
+base_model = "alibaba/qwen3.6-27b"
+reasoning_options = []
+
+[cost]
+input = 0.30
+output = 2.00

providers/vultr/models/XiaomiMiMo/MiMo-V2.5-Pro.toml

@@ -0,0 +1,6 @@
+base_model = "xiaomi/mimo-v2.5-pro"
+reasoning_options = []
+
+[cost]
+input = 0.55
+output = 1.65

providers/vultr/models/deepseek-ai/DeepSeek-V4-Flash.toml

@@ -0,0 +1,6 @@
+base_model = "deepseek/deepseek-v4-flash"
+reasoning_options = []
+
+[cost]
+input = 0.30
+output = 1.00

providers/vultr/models/moonshotai/Kimi-K2.6.toml

@@ -5,8 +5,8 @@ reasoning_options = []
 field = "reasoning_content"
 
 [cost]
-input = 0.15
-output = 0.6
+input = 0.30
+output = 1.20
 
 [limit]
 output = 131_072

providers/vultr/models/nvidia/Llama-3.1-Nemotron-Safety-Guard-8B-v3.toml

@@ -1,23 +0,0 @@
-name = "Llama 3.1 Nemotron Safety Guard"
-base_model = "nvidia/llama-3.1-nemotron-safety-guard-8b-v3"
-family = "llama"
-attachment = false
-reasoning = false
-tool_call = false
-temperature = true
-open_weights = true
-knowledge = "2023-12"
-release_date = "2025-10-28"
-last_updated = "2025-10-28"
-
-[cost]
-input = 0.01
-output = 0.01
-
-[limit]
-context = 8_192
-output = 4_096
-
-[modalities]
-input = ["text"]
-output = ["text"]

providers/vultr/models/zai-org/GLM-5.2-FP8.toml

@@ -1,4 +1,4 @@
-base_model = "zhipuai/glm-5.1"
+base_model = "zhipuai/glm-5.2"
 reasoning_options = []
 
 [interleaved]
@@ -6,4 +6,7 @@ field = "reasoning_content"
 
 [cost]
 input = 0.85
-output = 3.1
+output = 3.10
+
+[limit]
+context = 393_216