QwenLM · Manouchehri · Aug 8, 2026 · Aug 8, 2026 · Aug 8, 2026 · Aug 8, 2026
diff --git a/docs/users/configuration/auth.md b/docs/users/configuration/auth.md
@@ -157,12 +157,28 @@ The key concept is **Model Providers** (`modelProviders`): Qwen Code supports mu
 
 #### Supported protocols
 
-| Protocol          | `modelProviders` key | Environment variables                                                                                | Providers                                                                                             |
-| ----------------- | -------------------- | ---------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
-| OpenAI-compatible | `openai`             | `OPENAI_API_KEY`, `OPENAI_BASE_URL`, `OPENAI_MODEL` (alias: `QWEN_MODEL`)                            | OpenAI, Azure OpenAI, OpenRouter, Requesty, ModelScope, Alibaba Cloud, any OpenAI-compatible endpoint |
-| Anthropic         | `anthropic`          | `ANTHROPIC_API_KEY`, `ANTHROPIC_BASE_URL`, `ANTHROPIC_MODEL`                                         | Anthropic Claude                                                                                      |
-| Google GenAI      | `gemini`             | `GEMINI_API_KEY`, `GEMINI_MODEL`                                                                     | Google Gemini                                                                                         |
-| Vertex AI         | `vertex-ai`          | `GOOGLE_API_KEY`, `GOOGLE_MODEL` (sets `GOOGLE_GENAI_USE_VERTEXAI=true`; uses the `gemini` protocol) | Google Vertex AI                                                                                      |
+| Protocol           | `modelProviders` key | Environment variables                                                                                | Providers                                                                                             |
+| ------------------ | -------------------- | ---------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- |
+| OpenAI-compatible  | `openai`             | `OPENAI_API_KEY`, `OPENAI_BASE_URL`, `OPENAI_MODEL` (alias: `QWEN_MODEL`)                            | OpenAI, Azure OpenAI, OpenRouter, Requesty, ModelScope, Alibaba Cloud, any OpenAI-compatible endpoint |
+| Anthropic          | `anthropic`          | `ANTHROPIC_API_KEY`, `ANTHROPIC_BASE_URL`, `ANTHROPIC_MODEL`                                         | Anthropic Claude                                                                                      |
+| Google GenAI       | `gemini`             | `GEMINI_API_KEY`, `GEMINI_MODEL`                                                                     | Google Gemini                                                                                         |
+| Vertex AI          | `vertex-ai`          | `GOOGLE_API_KEY`, `GOOGLE_MODEL` (sets `GOOGLE_GENAI_USE_VERTEXAI=true`; uses the `gemini` protocol) | Google Vertex AI                                                                                      |
+| DashScope (native) | `dashscope`          | `DASHSCOPE_API_KEY`, `DASHSCOPE_BASE_URL`, `DASHSCOPE_MODEL`                                         | Alibaba ModelStudio's native DashScope API (`qwen3.8-max`)                                            |
+
+#### DashScope (native API)
+
+`dashscope` connects to Alibaba ModelStudio's native DashScope generation API instead of
+the OpenAI-compatible endpoint used by the `openai` protocol above. Get an API key from
+[Alibaba Cloud ModelStudio](https://modelstudio.console.alibabacloud.com/), then either:
+
+- Run `/auth`, select **Alibaba ModelStudio**, then choose **Native DashScope API** from
+  the sub-menu, or
+- Set `DASHSCOPE_API_KEY` and run `qwen --auth-type=dashscope`.
+
+`dashscope` uses the exact same API key as the ModelStudio **Standard API Key** option
+(the `openai`-protocol, OpenAI-compatible setup) — only the wire protocol differs. See
+[Alibaba ModelStudio — Native DashScope API](model-providers.md#alibaba-modelstudio--native-dashscope-api-dashscope-protocol)
+for prompt caching, thinking control, and full `modelProviders` configuration details.
 
 #### Step 1: Configure models and providers in `~/.qwen/settings.json`
 

diff --git a/docs/users/configuration/model-providers.md b/docs/users/configuration/model-providers.md
@@ -26,13 +26,14 @@ Below are comprehensive configuration examples for different authentication type
 
 The `modelProviders` object keys must be valid `authType` values. Currently supported auth types are:
 
-| Auth Type    | Description                                                                                                                                     |
-| ------------ | ----------------------------------------------------------------------------------------------------------------------------------------------- |
-| `openai`     | OpenAI-compatible APIs (OpenAI, Azure OpenAI, local inference servers like vLLM/Ollama)                                                         |
-| `anthropic`  | Anthropic Claude API                                                                                                                            |
-| `gemini`     | Google Gemini API                                                                                                                               |
-| `qwen-oauth` | Qwen OAuth (hard-coded, cannot be overridden in `modelProviders`)                                                                               |
-| `vertex-ai`  | Google Vertex AI (uses the `gemini` protocol and the `@google/genai` SDK in Vertex AI mode; selecting it sets `GOOGLE_GENAI_USE_VERTEXAI=true`) |
+| Auth Type    | Description                                                                                                                                                                                           |
+| ------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `openai`     | OpenAI-compatible APIs (OpenAI, Azure OpenAI, local inference servers like vLLM/Ollama)                                                                                                               |
+| `anthropic`  | Anthropic Claude API                                                                                                                                                                                  |
+| `gemini`     | Google Gemini API                                                                                                                                                                                     |
+| `qwen-oauth` | Qwen OAuth (hard-coded, cannot be overridden in `modelProviders`)                                                                                                                                     |
+| `vertex-ai`  | Google Vertex AI (uses the `gemini` protocol and the `@google/genai` SDK in Vertex AI mode; selecting it sets `GOOGLE_GENAI_USE_VERTEXAI=true`)                                                       |
+| `dashscope`  | Alibaba ModelStudio's native DashScope generation API (not OpenAI-compatible) — see [Alibaba ModelStudio — Native DashScope API](#alibaba-modelstudio--native-dashscope-api-dashscope-protocol) below |
 
 > [!warning]
 > A provider id that is neither a built-in protocol nor mapped via `providerProtocol` (e.g. a typo like `"openai-custom"`) cannot be routed, so its whole entry is **skipped** with a warning — its models simply won't appear in the `/model` picker. Use one of the supported auth type values above for built-in providers, or add a [`providerProtocol`](#custom-provider-ids-providerprotocol) mapping for a custom id.
@@ -70,6 +71,7 @@ Qwen Code uses the following official SDKs to send requests to each provider:
 | `anthropic`  | [`@anthropic-ai/sdk`](https://www.npmjs.com/package/@anthropic-ai/sdk) - Official Anthropic SDK |
 | `gemini`     | [`@google/genai`](https://www.npmjs.com/package/@google/genai) - Official Google GenAI SDK      |
 | `qwen-oauth` | [`openai`](https://www.npmjs.com/package/openai) with custom provider (DashScope-compatible)    |
+| `dashscope`  | Built-in `fetch`-based transport speaking DashScope's native wire format directly (no SDK)      |
 
 This means the `baseUrl` you configure should be compatible with the corresponding SDK's expected API format. For example, when using `openai` auth type, the endpoint must accept OpenAI API format requests.
 
@@ -204,6 +206,94 @@ This auth type supports not only OpenAI's official API but also any OpenAI-compa
 }
 ```
 
+### Alibaba ModelStudio — Native DashScope API (`dashscope` protocol)
+
+`dashscope` is a separate auth type from `openai`: instead of going through the
+OpenAI-compatible `/compatible-mode/v1` endpoint, it speaks Alibaba ModelStudio's
+**native** DashScope generation API
+(`/api/v1/services/aigc/multimodal-generation/generation`) directly. It currently
+supports the `qwen3.8-max` model.
+
+Prefer `dashscope` over the OpenAI-compatible `openai` protocol (with a DashScope
+`baseUrl`) when you want:
+
+- **Explicit prompt caching** with separate cache-read and cache-creation token counts
+  reported in usage — cached input is billed at roughly 0.1x, cache writes at roughly
+  1.25x. Cache markers are placed automatically; disable them per model with
+  `generationConfig.enableCacheControl: false`.
+- **Raw `reasoning_content` streaming** — thinking output arrives in its own field, with
+  no `<think>` tag scraping.
+- **Native parallel tool calls.**
+
+Both `dashscope` and the OpenAI-compatible `openai` protocol (pointed at a DashScope
+`baseUrl`) read the same `DASHSCOPE_API_KEY` — only the wire protocol differs.
+
+#### Regions
+
+| Region              | `baseUrl`                                    |
+| ------------------- | -------------------------------------------- |
+| Singapore (default) | `https://dashscope-intl.aliyuncs.com/api/v1` |
+| Beijing             | `https://dashscope.aliyuncs.com/api/v1`      |
+| US (Virginia)       | `https://dashscope-us.aliyuncs.com/api/v1`   |
+
+#### Settings example
+
+```json
+{
+  "security": { "auth": { "selectedType": "dashscope" } },
+  "model": { "name": "qwen3.8-max" },
+  "modelProviders": {
+    "dashscope": [
+      {
+        "id": "qwen3.8-max",
+        "name": "[ModelStudio Native] qwen3.8-max",
+        "baseUrl": "https://dashscope-intl.aliyuncs.com/api/v1",
+        "envKey": "DASHSCOPE_API_KEY",
+        "generationConfig": {
+          "contextWindowSize": 1000000,
+          "reasoning": { "effort": "medium" }
+        }
+      }
+    ]
+  }
+}
+```
+
+#### Headless / CI recipe
+
+```bash
+export DASHSCOPE_API_KEY="sk-..."
+export DASHSCOPE_MODEL="qwen3.8-max"
+qwen --auth-type=dashscope -p "your prompt"
+```
+
+Optionally set `DASHSCOPE_BASE_URL` to pick a different region than the Singapore
+default.
+
+#### Thinking control
+
+`generationConfig.reasoning` maps to the native `reasoning_effort` parameter:
+
+- `reasoning: { "effort": "low" | "medium" | "xhigh" }` sets
+  `reasoning_effort` to that native tier. Effort controls reject `high` and
+  `max`; if either remains in persisted configuration, request conversion
+  clamps it to `xhigh`.
+- `reasoning: false` disables thinking (`reasoning_effort: "none"`).
+
+Native-only knobs not covered by `reasoning` — e.g. `thinking_budget`,
+`max_completion_tokens`, `enable_search` — can be passed through via
+`generationConfig.extra_body`.
+
+#### Caveats
+
+- Forcing a tool choice (function-calling mode `ANY` / a specific function name) disables
+  thinking for that request — this is a native DashScope API constraint, not a qwen-code
+  limitation.
+- Embeddings are not supported on this protocol; use the OpenAI-compatible `openai`
+  protocol against a DashScope embeddings endpoint instead.
+- Multimodal input currently supports images. Video/PDF/audio content blocks are passed
+  through but not yet battle-tested.
+
 ### Google Gemini (`gemini`)
 
 ```json

diff --git a/packages/cli/src/acp-integration/acpAgent.ts b/packages/cli/src/acp-integration/acpAgent.ts
@@ -916,6 +916,7 @@ export const AUTH_PREFLIGHT_ENV_KEYS: Readonly<
   anthropic: ['ANTHROPIC_API_KEY'],
   gemini: ['GEMINI_API_KEY'],
   'vertex-ai': ['GOOGLE_API_KEY'],
+  dashscope: ['DASHSCOPE_API_KEY'],
 };
 
 /**

diff --git a/packages/cli/src/acp-integration/authMethods.test.ts b/packages/cli/src/acp-integration/authMethods.test.ts
@@ -17,6 +17,7 @@ describe('ACP auth methods', () => {
 
     expect(authMethods.map((method) => method.id)).toEqual([
       AuthType.USE_OPENAI,
+      AuthType.USE_DASHSCOPE,
     ]);
   });
 
@@ -25,6 +26,7 @@ describe('ACP auth methods', () => {
 
     expect(authMethods.map((method) => method.id)).toEqual([
       AuthType.USE_OPENAI,
+      AuthType.USE_DASHSCOPE,
     ]);
   });
 });
diff --git a/packages/cli/src/acp-integration/authMethods.ts b/packages/cli/src/acp-integration/authMethods.ts
@@ -18,6 +18,13 @@ export function buildAuthMethods(): AuthMethod[] {
         args: ['--auth-type=openai'],
       },
     },
+    {
+      id: AuthType.USE_DASHSCOPE,
+      name: 'Use DashScope API key (native)',
+      description:
+        'Requires setting the `DASHSCOPE_API_KEY` environment variable',
+      _meta: { type: 'terminal', args: ['--auth-type=dashscope'] },
+    },
   ];
 }
 

diff --git a/packages/cli/src/config/auth.test.ts b/packages/cli/src/config/auth.test.ts
@@ -36,6 +36,7 @@ describe('validateAuthMethod', () => {
     delete process.env['GOOGLE_API_KEY'];
     delete process.env['IDEALAB_KEY'];
     delete process.env['TOKEN_PLAN_KEY'];
+    delete process.env['DASHSCOPE_API_KEY'];
   });
 
   it('should return null for USE_OPENAI with default env key', () => {
@@ -49,6 +50,17 @@ describe('validateAuthMethod', () => {
     );
   });
 
+  it('should return null for USE_DASHSCOPE with default env key', () => {
+    process.env['DASHSCOPE_API_KEY'] = 'fake-key';
+    expect(validateAuthMethod(AuthType.USE_DASHSCOPE)).toBeNull();
+  });
+
+  it('should return an error message for USE_DASHSCOPE if no API key is available', () => {
+    expect(validateAuthMethod(AuthType.USE_DASHSCOPE)).toBe(
+      "Missing API key for DashScope native auth. Set settings.security.auth.apiKey, or set the 'DASHSCOPE_API_KEY' environment variable.",
+    );
+  });
+
   it('should return null for USE_OPENAI with custom envKey from modelProviders', () => {
     vi.mocked(settings.loadSettings).mockReturnValue({
       merged: {

diff --git a/packages/cli/src/config/auth.ts b/packages/cli/src/config/auth.ts
@@ -23,6 +23,7 @@ const DEFAULT_ENV_KEYS: Record<string, string> = {
   [AuthType.USE_ANTHROPIC]: 'ANTHROPIC_API_KEY',
   [AuthType.USE_GEMINI]: 'GEMINI_API_KEY',
   [AuthType.USE_VERTEX_AI]: 'GOOGLE_API_KEY',
+  [AuthType.USE_DASHSCOPE]: 'DASHSCOPE_API_KEY',
 };
 
 /**
@@ -259,6 +260,32 @@ export function validateAuthMethod(
     return null;
   }
 
+  if (authMethod === AuthType.USE_DASHSCOPE) {
+    const { hasKey, checkedEnvKey, isExplicitEnvKey } = hasApiKeyForAuth(
+      authMethod,
+      settings.merged,
+      config,
+    );
+    if (!hasKey) {
+      const envKeyHint = checkedEnvKey
+        ? `'${checkedEnvKey}'`
+        : "'DASHSCOPE_API_KEY'";
+      if (isExplicitEnvKey) {
+        // Explicit envKey configured - only suggest setting the env var
+        return t(
+          'Missing API key for DashScope native auth. Set the {{envKeyHint}} environment variable.',
+          { envKeyHint },
+        );
+      }
+      // Default env key - can use either apiKey or env var
+      return t(
+        'Missing API key for DashScope native auth. Set settings.security.auth.apiKey, or set the {{envKeyHint}} environment variable.',
+        { envKeyHint },
+      );
+    }
+    return null;
+  }
+
   if (authMethod === AuthType.QWEN_OAUTH) {
     // Qwen OAuth free tier was discontinued on 2026-04-15.
     // Block new OAuth setups; existing cached tokens still work until server rejects them.

diff --git a/packages/cli/src/config/config.ts b/packages/cli/src/config/config.ts
@@ -955,6 +955,7 @@ export async function parseArguments(): Promise<CliArgs> {
             AuthType.QWEN_OAUTH,
             AuthType.USE_GEMINI,
             AuthType.USE_VERTEX_AI,
+            AuthType.USE_DASHSCOPE,
           ],
           description: 'Authentication type',
         })

diff --git a/packages/cli/src/config/settingsSchema.test.ts b/packages/cli/src/config/settingsSchema.test.ts
@@ -435,6 +435,9 @@ describe('SettingsSchema', () => {
       expect(getSettingsSchema().permissions.showInDialog).toBe(false);
       expect(getSettingsSchema().mcpServers.showInDialog).toBe(false);
       expect(getSettingsSchema().telemetry.showInDialog).toBe(false);
+      expect(
+        getSettingsSchema().model.properties.reasoningEffort.showInDialog,
+      ).toBe(true);
 
       // Check that some settings are appropriately hidden
       expect(getSettingsSchema().ui.properties.theme.showInDialog).toBe(true);

diff --git a/packages/cli/src/config/settingsSchema.ts b/packages/cli/src/config/settingsSchema.ts
@@ -375,7 +375,7 @@ const SETTINGS_SCHEMA = {
     requiresRestart: true,
     default: {} as ProviderProtocolConfig,
     description:
-      'Maps a custom modelProviders provider id to the SDK protocol that routes its requests (e.g. {"idealab": "openai"}). Lets a custom provider id reuse a built-in protocol. Built-in provider ids (openai, gemini, anthropic, vertex-ai, qwen-oauth) are routed automatically and need no entry.',
+      'Maps a custom modelProviders provider id to the SDK protocol that routes its requests (e.g. {"idealab": "openai"}). Lets a custom provider id reuse a built-in protocol. Built-in provider ids (openai, gemini, anthropic, vertex-ai, qwen-oauth, dashscope) are routed automatically and need no entry.',
     showInDialog: false,
     mergeStrategy: MergeStrategy.REPLACE,
   },

diff --git a/packages/cli/src/i18n/locales/ca.js b/packages/cli/src/i18n/locales/ca.js
@@ -1969,8 +1969,8 @@ export default {
     '(Useu les fletxes ↑ ↓ per navegar, Return per seleccionar, Ctrl+C per sortir)\n',
   'Switch to plan mode or exit plan mode':
     'Canviar al mode de planificació o sortir del mode de planificació',
-  'Set how hard reasoning-capable models think ({{tiers}}); mapped and clamped per provider.':
-    'Defineix la intensitat de raonament dels models compatibles ({{tiers}}); mapejada i limitada segons el proveïdor.',
+  'Set how hard reasoning-capable models think; available tiers depend on the active provider/model.':
+    'Defineix la intensitat de raonament dels models compatibles; els nivells disponibles depenen del proveïdor/model actiu.',
   'Exited plan mode. Previous approval mode restored.':
     "S'ha sortit del mode de planificació. S'ha restaurat el mode d'aprovació anterior.",
   'Enabled plan mode. The agent will analyze and plan without executing tools.':

diff --git a/packages/cli/src/i18n/locales/de.js b/packages/cli/src/i18n/locales/de.js
@@ -1928,8 +1928,8 @@ export default {
   'to expand details': 'zum Erweitern der Details',
   'Switch to plan mode or exit plan mode':
     'In den Plan-Modus wechseln oder den Plan-Modus verlassen',
-  'Set how hard reasoning-capable models think ({{tiers}}); mapped and clamped per provider.':
-    'Legt fest, wie intensiv schlussfolgerungsfähige Modelle denken ({{tiers}}); je nach Anbieter zugeordnet und begrenzt.',
+  'Set how hard reasoning-capable models think; available tiers depend on the active provider/model.':
+    'Legt fest, wie intensiv schlussfolgerungsfähige Modelle denken; verfügbare Stufen hängen vom aktiven Anbieter/Modell ab.',
   'Exited plan mode. Previous approval mode restored.':
     'Plan-Modus verlassen. Vorheriger Genehmigungsmodus wiederhergestellt.',
   'Enabled plan mode. The agent will analyze and plan without executing tools.':

diff --git a/packages/cli/src/i18n/locales/en.js b/packages/cli/src/i18n/locales/en.js
@@ -2488,8 +2488,8 @@ export default {
     '(Use ↑ ↓ arrows to navigate, Enter to select, Ctrl+C to exit)\n',
   'Switch to plan mode or exit plan mode':
     'Switch to plan mode or exit plan mode',
-  'Set how hard reasoning-capable models think ({{tiers}}); mapped and clamped per provider.':
-    'Set how hard reasoning-capable models think ({{tiers}}); mapped and clamped per provider.',
+  'Set how hard reasoning-capable models think; available tiers depend on the active provider/model.':
+    'Set how hard reasoning-capable models think; available tiers depend on the active provider/model.',
   'Set a goal — keep working until the condition is met':
     'Set a goal — keep working until the condition is met',
   'Set or control a session goal': 'Set or control a session goal',