feat(minimax): add MiniMax provider with M3 as default

docs/users/configuration/model-providers.md

@@ -195,6 +195,71 @@ This auth type supports not only OpenAI's official API but also any OpenAI-compa
 }
 ```
 
+### MiniMax (`openai`)
+
+MiniMax provides OpenAI-compatible chat models. Configure using the `openai` auth type with MiniMax's base URL. MiniMax requires temperature to be in the range `(0.0, 1.0]` — the provider enforces this automatically.
+
+| Model ID                 | Description                                                                    |
+| ------------------------ | ------------------------------------------------------------------------------ |
+| `MiniMax-M3`             | Latest model (default) — 512K context, up to 128K output, supports image input |
+| `MiniMax-M2.7`           | Peak performance, ultimate value — handles complex tasks                       |
+| `MiniMax-M2.7-highspeed` | Same performance, faster and more agile                                        |
+
+```json
+{
+  "env": {
+    "MINIMAX_API_KEY": "your-minimax-api-key"
+  },
+  "modelProviders": {
+    "openai": [
+      {
+        "id": "MiniMax-M3",
+        "name": "MiniMax-M3",
+        "envKey": "MINIMAX_API_KEY",
+        "baseUrl": "https://api.minimax.io/v1",
+        "generationConfig": {
+          "contextWindowSize": 524288,
+          "samplingParams": {
+            "temperature": 1.0,
+            "max_tokens": 131072
+          }
+        }
+      },
+      {
+        "id": "MiniMax-M2.7",
+        "name": "MiniMax-M2.7",
+        "envKey": "MINIMAX_API_KEY",
+        "baseUrl": "https://api.minimax.io/v1",
+        "generationConfig": {
+          "contextWindowSize": 200000,
+          "samplingParams": {
+            "temperature": 1.0,
+            "max_tokens": 65536
+          }
+        }
+      },
+      {
+        "id": "MiniMax-M2.7-highspeed",
+        "name": "MiniMax-M2.7-highspeed",
+        "envKey": "MINIMAX_API_KEY",
+        "baseUrl": "https://api.minimax.io/v1",
+        "generationConfig": {
+          "contextWindowSize": 200000,
+          "samplingParams": {
+            "temperature": 1.0,
+            "max_tokens": 65536
+          }
+        }
+      }
+    ]
+  }
+}
+```
+
+> [!note]
+>
+> **Domestic mirror**: Users in mainland China can replace `api.minimax.io` with `api.minimaxi.com` (note the extra `i`).
+
 ### Local Self-Hosted Models (via OpenAI-compatible API)
 
 Most local inference servers (vLLM, Ollama, LM Studio, etc.) provide an OpenAI-compatible API endpoint. Configure them using the `openai` auth type with a local `baseUrl`:

packages/core/src/core/modalityDefaults.test.ts

@@ -171,8 +171,16 @@ describe('defaultModalities', () => {
   });
 
   describe('MiniMax', () => {
-    it('returns text-only for MiniMax-M2.5', () => {
-      expect(defaultModalities('MiniMax-M2.5')).toEqual({});
+    it('returns image for MiniMax-M3', () => {
+      const m = defaultModalities('MiniMax-M3');
+      expect(m.image).toBe(true);
+      expect(m.pdf).toBeUndefined();
+      expect(m.audio).toBeUndefined();
+      expect(m.video).toBeUndefined();
+    });
+
+    it('returns text-only for MiniMax-M2.7', () => {
+      expect(defaultModalities('MiniMax-M2.7')).toEqual({});
     });
   });
 

packages/core/src/core/modalityDefaults.ts

@@ -66,8 +66,11 @@ const MODALITY_PATTERNS: Array<[RegExp, InputModalities]> = [
   [/^glm-/, {}],
 
   // -------------------
-  // MiniMax — text-only
+  // MiniMax
   // -------------------
+  // M3 supports image input (images only, no video/audio/documents)
+  [/^minimax-m3/, { image: true }],
+  // M2.7 and earlier: text-only
   [/^minimax-/, {}],
 
   // -------------------

packages/core/src/core/openaiContentGenerator/constants.ts

@@ -6,3 +6,4 @@ export const DEFAULT_DASHSCOPE_BASE_URL =
   'https://dashscope.aliyuncs.com/compatible-mode/v1';
 export const DEFAULT_DEEPSEEK_BASE_URL = 'https://api.deepseek.com/v1';
 export const DEFAULT_OPEN_ROUTER_BASE_URL = 'https://openrouter.ai/api/v1';
+export const DEFAULT_MINIMAX_BASE_URL = 'https://api.minimax.io/v1';

packages/core/src/core/openaiContentGenerator/index.ts

@@ -13,6 +13,7 @@ import { OpenAIContentGenerator } from './openaiContentGenerator.js';
 import {
   DashScopeOpenAICompatibleProvider,
   DeepSeekOpenAICompatibleProvider,
+  MiniMaxOpenAICompatibleProvider,
   ModelScopeOpenAICompatibleProvider,
   OpenRouterOpenAICompatibleProvider,
   type OpenAICompatibleProvider,
@@ -26,6 +27,7 @@ export {
   type OpenAICompatibleProvider,
   DashScopeOpenAICompatibleProvider,
   DeepSeekOpenAICompatibleProvider,
+  MiniMaxOpenAICompatibleProvider,
   OpenRouterOpenAICompatibleProvider,
 } from './provider/index.js';
 
@@ -87,6 +89,14 @@ export function determineProvider(
     );
   }
 
+  // Check for MiniMax provider
+  if (MiniMaxOpenAICompatibleProvider.isMiniMaxProvider(config)) {
+    return new MiniMaxOpenAICompatibleProvider(
+      contentGeneratorConfig,
+      cliConfig,
+    );
+  }
+
   // Default provider for standard OpenAI-compatible APIs
   return new DefaultOpenAICompatibleProvider(contentGeneratorConfig, cliConfig);
 }

packages/core/src/core/openaiContentGenerator/provider/index.ts

@@ -2,6 +2,7 @@ export { ModelScopeOpenAICompatibleProvider } from './modelscope.js';
 export { DashScopeOpenAICompatibleProvider } from './dashscope.js';
 export { DeepSeekOpenAICompatibleProvider } from './deepseek.js';
 export { OpenRouterOpenAICompatibleProvider } from './openrouter.js';
+export { MiniMaxOpenAICompatibleProvider } from './minimax.js';
 export { DefaultOpenAICompatibleProvider } from './default.js';
 export type {
   OpenAICompatibleProvider,