feat: add new DeepInfra and SiliconFlow models to model list

litellm/config.yaml

@@ -50,7 +50,35 @@ model_list:
       api_key: os.environ/SILICONFLOW_API_KEY
 
   # ═══════════════════════════════════════════════
-  # TIER 3: Other DeepInfra models
+  # TIER 3: Ultra-cheap DeepInfra models
+  # ═══════════════════════════════════════════════
+
+  # GPT-OSS-120B — OpenAI open-weight MoE ($0.05 in / $0.24 out per M)
+  - model_name: gpt-oss
+    litellm_params:
+      model: deepinfra/openai/gpt-oss-120b
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # GPT-OSS-20B — lower latency variant ($0.04 in / $0.16 out per M)
+  - model_name: gpt-oss-20b
+    litellm_params:
+      model: deepinfra/openai/gpt-oss-20b
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # Nemotron Super 49B — near-flagship quality ($0.10 in / $0.40 out per M)
+  - model_name: nemotron-super
+    litellm_params:
+      model: deepinfra/nvidia/Llama-3.3-Nemotron-Super-49B-v1.5
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # Nemotron Nano 9B — dirt cheap for simple tasks ($0.04 in / $0.16 out per M)
+  - model_name: nemotron-nano
+    litellm_params:
+      model: deepinfra/nvidia/NVIDIA-Nemotron-Nano-9B-v2
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # ═══════════════════════════════════════════════
+  # TIER 4: Other DeepInfra models
   # ═══════════════════════════════════════════════
 
   - model_name: deepseek-r1
@@ -64,16 +92,55 @@ model_list:
       api_key: os.environ/DEEPINFRA_API_KEY
 
   # ═══════════════════════════════════════════════
-  # TIER 4: SiliconFlow (Qwen/GLM)
+  # TIER 5: GLM models (cheapest first)
   # ═══════════════════════════════════════════════
 
-  # GLM-4.7 via SiliconFlow
+  # GLM-4.6 via DeepInfra ($0.60 in / $1.90 out per M)
+  - model_name: glm-4.6
+    litellm_params:
+      model: deepinfra/zai-org/GLM-4.6
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # GLM-4.7 via DeepInfra ($0.40 in / $1.75 out per M)
+  - model_name: glm-4.7
+    litellm_params:
+      model: deepinfra/zai-org/GLM-4.7
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # GLM-4.7 fallback via SiliconFlow
   - model_name: glm-4.7
     litellm_params:
       model: openai/THUDM/GLM-4-32B-0414
       api_base: https://api.siliconflow.com/v1
       api_key: os.environ/SILICONFLOW_API_KEY
 
+  # GLM-5 via DeepInfra ($0.80 in / $2.56 out per M)
+  - model_name: glm-5
+    litellm_params:
+      model: deepinfra/zai-org/GLM-5
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # ═══════════════════════════════════════════════
+  # TIER 6: Kimi K2 (cheapest first)
+  # ═══════════════════════════════════════════════
+
+  # Kimi K2 via DeepInfra ($0.50 in / $2.00 out per M)
+  - model_name: kimi-k2
+    litellm_params:
+      model: deepinfra/moonshotai/Kimi-K2-Instruct-0905
+      api_key: os.environ/DEEPINFRA_API_KEY
+
+  # Kimi K2 fallback via SiliconFlow ($0.58 in / $2.29 out per M)
+  - model_name: kimi-k2
+    litellm_params:
+      model: openai/moonshotai/Kimi-K2-Instruct-0905
+      api_base: https://api.siliconflow.com/v1
+      api_key: os.environ/SILICONFLOW_API_KEY
+
+  # ═══════════════════════════════════════════════
+  # TIER 7: SiliconFlow (Qwen)
+  # ═══════════════════════════════════════════════
+
   # Qwen3 Coder 480B MoE via SiliconFlow ($1.14 in / $2.28 out per M)
   - model_name: qwen3-coder
     litellm_params:
@@ -89,7 +156,7 @@ model_list:
       api_key: os.environ/SILICONFLOW_API_KEY
 
   # ═══════════════════════════════════════════════
-  # TIER 5: OpenRouter (most expensive, widest selection)
+  # TIER 8: OpenRouter (most expensive, widest selection)
   # ═══════════════════════════════════════════════
 
   # Kimi K2.5 — DeepInfra is cheapest ($0.45 in / $2.25 out per M)
@@ -109,7 +176,7 @@ model_list:
       model: openrouter/minimax/minimax-m2.5
       api_key: os.environ/OPENROUTER_API_KEY
 
-  - model_name: gpt-oss
+  - model_name: gpt-4.1-mini
     litellm_params:
       model: openrouter/openai/gpt-4.1-mini
       api_key: os.environ/OPENROUTER_API_KEY