feat: upgrade MiniMax model references to M3

examples/beginner/minimax_example.py

@@ -1,10 +1,10 @@
 """MiniMax AI Research Agent
 
-A Bindu agent powered by MiniMax's M2.7 model via OpenAI-compatible API.
+A Bindu agent powered by MiniMax's M3 model via OpenAI-compatible API.
 MiniMax offers high-performance models with up to 1M context window.
 
 Features:
-- MiniMax M2.7 model (1M context)
+- MiniMax M3 model (1M context)
 - Web search integration via DuckDuckGo
 - Research and summarization capabilities
 
@@ -30,11 +30,11 @@
 MINIMAX_API_KEY = os.getenv("MINIMAX_API_KEY")
 MINIMAX_BASE_URL = "https://api.minimax.io/v1"
 
-# Define your agent with MiniMax M2.7
+# Define your agent with MiniMax M3
 agent = Agent(
     instructions="You are a research assistant that finds and summarizes information.",
     model=OpenAILike(
-        id="MiniMax-M2.7",
+        id="MiniMax-M3",
         api_key=MINIMAX_API_KEY,
         base_url=MINIMAX_BASE_URL,
     ),

gateway/.env.example

@@ -35,7 +35,7 @@ OPENROUTER_API_KEY=sk-or-v1-...
 # OpenRouter tries each fallback in order. No per-call retry code
 # needed — OpenRouter handles it transparently.
 #
-# OPENROUTER_FALLBACK_MODELS=minimax/minimax-m2.7,openai/gpt-4o-mini
+# OPENROUTER_FALLBACK_MODELS=minimax/minimax-m3,openai/gpt-4o-mini
 
 # Server
 GATEWAY_PORT=3774

gateway/src/config/loader.ts

@@ -126,7 +126,7 @@ function envOverrides(base: Record<string, any>): Record<string, any> {
 
   // Optional: comma-separated fallback model IDs. When primary fails
   // (rate limit, upstream error), OpenRouter tries each in order.
-  // Example: OPENROUTER_FALLBACK_MODELS="minimax/minimax-m2.7,openai/gpt-4o-mini"
+  // Example: OPENROUTER_FALLBACK_MODELS="minimax/minimax-m3,openai/gpt-4o-mini"
   if (process.env.OPENROUTER_FALLBACK_MODELS) {
     const fallbacks = process.env.OPENROUTER_FALLBACK_MODELS.split(",")
       .map((s) => s.trim())

gateway/tests/provider/openrouter-body-injection.test.ts

@@ -72,11 +72,11 @@ describe("injectFallbackModels — OpenRouter fail-over routing", () => {
       messages: [],
     })
     const out = JSON.parse(
-      injectFallbackModels(body, ["minimax/minimax-m2.7", "openai/gpt-4o-mini"]),
+      injectFallbackModels(body, ["minimax/minimax-m3", "openai/gpt-4o-mini"]),
     )
     expect(out.models).toEqual([
       "anthropic/claude-sonnet-4.6",
-      "minimax/minimax-m2.7",
+      "minimax/minimax-m3",
       "openai/gpt-4o-mini",
     ])
     expect(out.route).toBe("fallback")

tests/unit/test_minimax_example.py

@@ -55,10 +55,10 @@ def test_example_has_minimax_base_url(self):
         source = EXAMPLE_PATH.read_text()
         assert "https://api.minimax.io/v1" in source
 
-    def test_example_uses_minimax_m27(self):
-        """Verify the example uses MiniMax-M2.7 model."""
+    def test_example_uses_minimax_m3(self):
+        """Verify the example uses MiniMax-M3 model."""
         source = EXAMPLE_PATH.read_text()
-        assert "MiniMax-M2.7" in source
+        assert "MiniMax-M3" in source
 
     def test_example_reads_api_key_from_env(self):
         """Verify the example reads MINIMAX_API_KEY from env."""
@@ -125,16 +125,15 @@ class TestMiniMaxModelConstants:
     def test_valid_minimax_models(self):
         """Verify known MiniMax model identifiers."""
         valid_models = {
+            "MiniMax-M3",
             "MiniMax-M2.7",
             "MiniMax-M2.7-highspeed",
-            "MiniMax-M2.5",
-            "MiniMax-M2.5-highspeed",
         }
         # Parse the example to check the model used
         source = EXAMPLE_PATH.read_text()
         for model in valid_models:
-            # At least M2.7 should be referenced
-            if model == "MiniMax-M2.7":
+            # At least M3 should be referenced
+            if model == "MiniMax-M3":
                 assert model in source
 
     def test_minimax_api_url_format(self):
@@ -200,7 +199,7 @@ def test_minimax_api_connection(self, api_key):
                 "Content-Type": "application/json",
             },
             json={
-                "model": "MiniMax-M2.7",
+                "model": "MiniMax-M3",
                 "messages": [{"role": "user", "content": "Hi"}],
                 "max_tokens": 5,
             },
@@ -219,7 +218,7 @@ def test_minimax_chat_completion(self, api_key):
                 "Content-Type": "application/json",
             },
             json={
-                "model": "MiniMax-M2.7",
+                "model": "MiniMax-M3",
                 "messages": [{"role": "user", "content": "Say hello in one word."}],
                 "max_tokens": 10,
             },