feat(gateway): update priority

llm-gateway.yaml

@@ -59,65 +59,56 @@ providers:
 
 models:
   # ═══ TIER 1: Free (OpenRouter free models, $0) ═══
-  - name: "llama-3.3-70b-free"
+  # NOTE: Commented out — free models are heavily rate-limited upstream.
-    routes:
+  #       Uncomment if you want best-effort free access.
-      - provider: openrouter
+  # - name: "llama-3.3-70b-free"
-        model: "meta-llama/llama-3.3-70b-instruct:free"
+  #   routes:
-
+  #     - provider: openrouter
-  - name: "deepseek-r1-free"
+  #       model: "meta-llama/llama-3.3-70b-instruct:free"
-    routes:
+  # - name: "deepseek-r1-free"
-      - provider: openrouter
+  #   routes:
-        model: "deepseek/deepseek-r1-0528:free"
+  #     - provider: openrouter
-
+  #       model: "deepseek/deepseek-r1-0528:free"
-  - name: "gpt-oss-free"
+  # - name: "gpt-oss-free"
-    routes:
+  #   routes:
-      - provider: openrouter
+  #     - provider: openrouter
-        model: "openai/gpt-oss-120b:free"
+  #       model: "openai/gpt-oss-120b:free"
-
+  # - name: "gpt-oss-20b-free"
-  - name: "gpt-oss-20b-free"
+  #   routes:
-    routes:
+  #     - provider: openrouter
-      - provider: openrouter
+  #       model: "openai/gpt-oss-20b:free"
-        model: "openai/gpt-oss-20b:free"
+  # - name: "qwen3-coder-free"
-
+  #   routes:
-  - name: "qwen3-coder-free"
+  #     - provider: openrouter
-    routes:
+  #       model: "qwen/qwen3-coder:free"
-      - provider: openrouter
+  # - name: "qwen3-235b-free"
-        model: "qwen/qwen3-coder:free"
+  #   routes:
-
+  #     - provider: openrouter
-  - name: "qwen3-235b-free"
+  #       model: "qwen/qwen3-235b-a22b-thinking-2507"
-    routes:
+  # - name: "glm-4.5-air-free"
-      - provider: openrouter
+  #   routes:
-        model: "qwen/qwen3-235b-a22b-thinking-2507"
+  #     - provider: openrouter
-
+  #       model: "z-ai/glm-4.5-air:free"
-  - name: "glm-4.5-air-free"
+  # - name: "nemotron-nano-free"
-    routes:
+  #   routes:
-      - provider: openrouter
+  #     - provider: openrouter
-        model: "z-ai/glm-4.5-air:free"
+  #       model: "nvidia/nemotron-nano-9b-v2:free"
-
+  # - name: "trinity-large-free"
-  - name: "nemotron-nano-free"
+  #   routes:
-    routes:
+  #     - provider: openrouter
-      - provider: openrouter
+  #       model: "arcee-ai/trinity-large-preview:free"
-        model: "nvidia/nemotron-nano-9b-v2:free"
+  # - name: "mistral-small-free"
-
+  #   routes:
-  - name: "trinity-large-free"
+  #     - provider: openrouter
-    routes:
+  #       model: "mistralai/mistral-small-3.1-24b-instruct:free"
-      - provider: openrouter
+  # - name: "gemma-3-27b-free"
-        model: "arcee-ai/trinity-large-preview:free"
+  #   routes:
-
+  #     - provider: openrouter
-  - name: "mistral-small-free"
+  #       model: "google/gemma-3-27b-it:free"
-    routes:
+  # - name: "step-3.5-flash-free"
-      - provider: openrouter
+  #   routes:
-        model: "mistralai/mistral-small-3.1-24b-instruct:free"
+  #     - provider: openrouter
-
+  #       model: "stepfun/step-3.5-flash:free"
-  - name: "gemma-3-27b-free"
-    routes:
-      - provider: openrouter
-        model: "google/gemma-3-27b-it:free"
-
-  - name: "step-3.5-flash-free"
-    routes:
-      - provider: openrouter
-        model: "stepfun/step-3.5-flash:free"
 
   # ═══ TIER 2: Low cost (Groq, Cerebras — free tier with rate limits) ═══
   - name: "llama-3.3-70b"
@@ -126,8 +117,11 @@ models:
         model: "llama-3.3-70b-versatile"
         pricing: { input: 0.59, output: 0.79 }
       - provider: deepinfra
-        model: "meta-llama/Llama-3.3-70B-Instruct"
+        model: "meta-llama/Llama-3.3-70B-Instruct-Turbo"
         pricing: { input: 0.23, output: 0.40 }
+      - provider: cerebras
+        model: "llama-3.3-70b"
+        pricing: { input: 0.85, output: 1.20 }
 
   - name: "llama-3.1-8b"
     routes:
@@ -162,6 +156,26 @@ models:
         model: "openai/gpt-oss-20b"
         pricing: { input: 0.04, output: 0.16 }
 
+  - name: "llama-4-scout"
+    routes:
+      - provider: groq
+        model: "meta-llama/llama-4-scout-17b-16e-instruct"
+        pricing: { input: 0.11, output: 0.34 }
+
+  - name: "llama-4-maverick"
+    routes:
+      - provider: groq
+        model: "meta-llama/llama-4-maverick-17b-128e-instruct"
+        pricing: { input: 0.20, output: 0.60 }
+
+  - name: "qwen3-32b"
+    routes:
+      - provider: groq
+        model: "qwen/qwen3-32b"
+        pricing: { input: 0.29, output: 0.59 }
+      - provider: cerebras
+        model: "qwen-3-32b"
+
   # ═══ TIER 3: DeepSeek V3.2 (cheapest flagship) ═══
   - name: "deepseek-v3.2"
     routes:
@@ -192,22 +206,24 @@ models:
   - name: "deepseek-r1"
     routes:
       - provider: deepinfra
-        model: "deepseek-ai/DeepSeek-R1"
+        model: "deepseek-ai/DeepSeek-R1-0528"
-        pricing: { input: 0.40, output: 1.60 }
       - provider: openrouter
         model: "deepseek/deepseek-r1"
-        pricing: { input: 0.55, output: 2.19 }
 
-  - name: "deepseek-r1-distill-qwen-32b"
+  - name: "deepseek-r1-distill-llama-70b"
     routes:
       - provider: deepinfra
-        model: "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B"
+        model: "deepseek-ai/DeepSeek-R1-Distill-Llama-70B"
-        pricing: { input: 0.07, output: 0.16 }
 
-  - name: "devstral"
+  - name: "devstral-small"
     routes:
-      - provider: deepinfra
+      - provider: openrouter
-        model: "mistralai/Devstral-Small-2505"
+        model: "mistralai/devstral-small"
+
+  - name: "devstral-medium"
+    routes:
+      - provider: openrouter
+        model: "mistralai/devstral-medium"
 
   # ═══ TIER 6: GLM ═══
   - name: "glm-4.6"
@@ -236,6 +252,9 @@ models:
   # ═══ TIER 7: Kimi ═══
   - name: "kimi-k2"
     routes:
+      - provider: groq
+        model: "moonshotai/kimi-k2-instruct-0905"
+        pricing: { input: 1.00, output: 3.00 }
       - provider: deepinfra
         model: "moonshotai/Kimi-K2-Instruct-0905"
         pricing: { input: 0.50, output: 2.00 }