fix(deepseek): honor thinking controls for canonical Flash (#144469)

Recognize canonical DeepSeek Flash through the existing thinking adapter and publish its current provider-owned metadata. Preserve legacy selections, the Pro default, explicit overrides, and discovery ownership.

Real Gateway request proof covers thinking controls, reasoning replay, images, and restored selections. Separate catalog-owner controls and focused native tests cover metadata and unchanged ownership boundaries.

Addresses the canonical capability and thinking-control gap in #144378. Thanks @gluo88 for the report and canonical-model analysis.

Co-authored-by: Ayaan Zaidi <hi@obviy.us>

docs/providers/deepseek.md

@@ -80,8 +80,9 @@ available to that process (for example, in `~/.openclaw/.env` or via
 
 | Model ref                               | Name                                    | Input       | Context   | Max output | Notes                            |
 | --------------------------------------- | --------------------------------------- | ----------- | --------- | ---------- | -------------------------------- |
+| `deepseek/deepseek-flash`               | DeepSeek V4.1 Flash                     | text, image | 1,000,000 | 384,000    | Canonical Flash model            |
 | `deepseek/deepseek-v4-flash`            | DeepSeek V4 Flash                       | text        | 1,000,000 | 384,000    | Fast V4 thinking-capable surface |
-| `deepseek/deepseek-v4-pro`              | DeepSeek V4 Pro                         | text        | 1,000,000 | 384,000    | Default; strongest V4 model      |
+| `deepseek/deepseek-v4-pro`              | DeepSeek V4 Pro                         | text        | 1,000,000 | 384,000    | Onboarding default               |
 | `deepseek/deepseek-v4-flash-vision-exp` | DeepSeek V4 Flash Vision (Experimental) | text, image | 1,000,000 | 384,000    | Experimental image understanding |
 
 <Warning>
@@ -90,19 +91,24 @@ DeepSeek retired `deepseek-chat` and `deepseek-reasoner` on July 24, 2026 at
 to `deepseek/deepseek-v4-flash` or `deepseek/deepseek-v4-pro`.
 </Warning>
 
-OpenClaw's local costs are estimates. The vision model's bundled estimate uses
-DeepSeek's peak rates; its published off-peak rates are half those amounts.
+OpenClaw's local costs are estimates. Canonical Flash uses DeepSeek's peak
+rates: $0.30 per million input tokens, $1.20 per million output tokens, and
+$0.006 per million cached input tokens. Published off-peak rates are half those amounts.
+Legacy rows retain their earlier bundled metadata. DeepSeek still accepts
+`deepseek-v4-flash` and `deepseek-v4-flash-vision-exp`, routes them to V4.1 Flash,
+and bills them at current Flash rates. Existing explicit selections remain valid in OpenClaw.
 DeepSeek can change rates; its
 [Models & Pricing](https://api-docs.deepseek.com/quick_start/pricing/) page is
 authoritative for billing.
 
-For image inputs, select `deepseek/deepseek-v4-flash-vision-exp`. The regular
-Flash and Pro models are text-only. DeepSeek's experimental vision model accepts
+For image inputs, select `deepseek/deepseek-flash`. The legacy
+`deepseek/deepseek-v4-flash-vision-exp` selection also retains image support.
+Canonical Flash accepts
 PNG, JPEG, GIF, and WebP images through the same API and API key. See
 [DeepSeek vision](https://api-docs.deepseek.com/guides/vision) for image limits.
 
 <Tip>
-V4 models support DeepSeek's `thinking` control. OpenClaw also replays
+Canonical Flash and V4 models support DeepSeek's `thinking` control. OpenClaw also replays
 DeepSeek `reasoning_content` on follow-up turns so thinking sessions with tool
 calls can continue.
 Use `/think xhigh` or `/think max` with DeepSeek V4 models to request DeepSeek's
@@ -114,7 +120,7 @@ maximum `reasoning_effort`; both map to `"max"`.
 DeepSeek V4 thinking sessions require replayed assistant messages from a
 thinking-enabled turn to include `reasoning_content` on follow-up requests.
 OpenClaw's DeepSeek plugin backfills that field automatically, so normal
-multi-turn tool use works on `deepseek/deepseek-v4-flash`,
+multi-turn tool use works on `deepseek/deepseek-flash`, `deepseek/deepseek-v4-flash`,
 `deepseek/deepseek-v4-flash-vision-exp`, and `deepseek/deepseek-v4-pro` even when history came from another
 OpenAI-compatible provider (no native `reasoning_content`) or from a plain
 assistant message. No `/new` required after switching providers mid-session.
@@ -123,9 +129,11 @@ When thinking is disabled (including the UI **None** selection), OpenClaw
 sends `thinking: { type: "disabled" }` and strips replayed `reasoning_content`
 from outgoing history, keeping the session on the non-thinking DeepSeek path.
 
-Fresh onboarding selects the stronger `deepseek/deepseek-v4-pro` model. Use
-`deepseek/deepseek-v4-flash` when lower cost or latency matters more than
-maximum capability.
+Fresh onboarding selects `deepseek/deepseek-v4-pro`. To select canonical Flash:
+
+```bash
+openclaw models set deepseek/deepseek-flash
+```
 
 ## Live testing
 

extensions/deepseek/index.test.ts

@@ -170,6 +170,7 @@ describe("deepseek provider plugin", () => {
       "deepseek-v4-flash",
       "deepseek-v4-pro",
       "deepseek-v4-flash-vision-exp",
+      "deepseek-flash",
     ]);
     const flashModel = catalogProvider.models?.find((model) => model.id === "deepseek-v4-flash");
     expect(flashModel?.reasoning).toBe(true);
@@ -189,6 +190,11 @@ describe("deepseek provider plugin", () => {
         ]),
       ),
     ).toEqual({
+      "deepseek-flash": {
+        contextWindow: 1_000_000,
+        maxTokens: 384_000,
+        cost: { input: 0.3, output: 1.2, cacheRead: 0.006, cacheWrite: 0 },
+      },
       "deepseek-v4-flash": {
         contextWindow: 1_000_000,
         maxTokens: 384_000,
@@ -308,6 +314,7 @@ describe("deepseek provider plugin", () => {
     const expectedV4Levels = ["off", "minimal", "low", "medium", "high", "xhigh", "max"];
 
     for (const modelId of [
+      "deepseek-flash",
       "deepseek-v4-flash",
       "deepseek-v4-pro",
       "deepseek-v4-flash-vision-exp",
@@ -324,60 +331,62 @@ describe("deepseek provider plugin", () => {
     ).toBe(undefined);
   });
 
-  it.each(["deepseek-v4-flash", "deepseek-v4-pro", "deepseek-v4-flash-vision-exp"])(
-    "maps thinking levels to %s payload controls",
-    async (modelId) => {
-      let capturedPayload: Record<string, unknown> | undefined;
-      const baseStreamFn = (
-        _model: Model<"openai-completions">,
-        _context: Context,
-        options?: { onPayload?: (payload: unknown) => unknown },
-      ) => {
-        capturedPayload = {
-          model: "deepseek-v4-pro",
-          reasoning_effort: "high",
-        };
-        options?.onPayload?.(capturedPayload);
-        const stream = createAssistantMessageEventStream();
-        queueMicrotask(() => stream.end());
-        return stream;
+  it.each([
+    "deepseek-flash",
+    "deepseek-v4-flash",
+    "deepseek-v4-pro",
+    "deepseek-v4-flash-vision-exp",
+  ])("maps thinking levels to %s payload controls", async (modelId) => {
+    let capturedPayload: Record<string, unknown> | undefined;
+    const baseStreamFn = (
+      _model: Model<"openai-completions">,
+      _context: Context,
+      options?: { onPayload?: (payload: unknown) => unknown },
+    ) => {
+      capturedPayload = {
+        model: "deepseek-v4-pro",
+        reasoning_effort: "high",
       };
+      options?.onPayload?.(capturedPayload);
+      const stream = createAssistantMessageEventStream();
+      queueMicrotask(() => stream.end());
+      return stream;
+    };
 
-      const wrapThinkingOff = expectDefined(
-        createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "off"),
-        "DeepSeek thinking wrapper for off",
-      );
-      await wrapThinkingOff(
-        {
-          provider: "deepseek",
-          id: modelId,
-          api: "openai-completions",
-        } as never,
-        { messages: [] } as never,
-        {},
-      );
+    const wrapThinkingOff = expectDefined(
+      createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "off"),
+      "DeepSeek thinking wrapper for off",
+    );
+    await wrapThinkingOff(
+      {
+        provider: "deepseek",
+        id: modelId,
+        api: "openai-completions",
+      } as never,
+      { messages: [] } as never,
+      {},
+    );
 
-      expect(readThinking(capturedPayload)?.type).toBe("disabled");
-      expect(capturedPayload).not.toHaveProperty("reasoning_effort");
+    expect(readThinking(capturedPayload)?.type).toBe("disabled");
+    expect(capturedPayload).not.toHaveProperty("reasoning_effort");
 
-      const wrapThinkingXhigh = expectDefined(
-        createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "xhigh"),
-        "DeepSeek thinking wrapper for xhigh",
-      );
-      await wrapThinkingXhigh(
-        {
-          provider: "deepseek",
-          id: modelId,
-          api: "openai-completions",
-        } as never,
-        { messages: [] } as never,
-        {},
-      );
+    const wrapThinkingXhigh = expectDefined(
+      createDeepSeekV4ThinkingWrapper(baseStreamFn as never, "xhigh"),
+      "DeepSeek thinking wrapper for xhigh",
+    );
+    await wrapThinkingXhigh(
+      {
+        provider: "deepseek",
+        id: modelId,
+        api: "openai-completions",
+      } as never,
+      { messages: [] } as never,
+      {},
+    );
 
-      expect(readThinking(capturedPayload)?.type).toBe("enabled");
-      expect(capturedPayload?.reasoning_effort).toBe("max");
-    },
-  );
+    expect(readThinking(capturedPayload)?.type).toBe("enabled");
+    expect(capturedPayload?.reasoning_effort).toBe("max");
+  });
 
   it.each(["deepseek-v4-flash", "deepseek-v4-flash-vision-exp"])(
     "preserves replayed reasoning_content for %s",
@@ -406,40 +415,43 @@ describe("deepseek provider plugin", () => {
     },
   );
 
-  it("keeps image input in requests for the bundled vision model", () => {
-    const provider = buildDeepSeekProvider();
-    const entry = provider.models?.find((model) => model.id === "deepseek-v4-flash-vision-exp");
-    expect(entry).toMatchObject({ reasoning: true, input: ["text", "image"] });
-    const model = {
-      ...entry,
-      provider: "deepseek",
-      baseUrl: provider.baseUrl,
-      api: "openai-completions",
-    } as OpenAICompletionsModel;
-    const payload = buildOpenAICompletionsParams(
-      model,
-      {
-        messages: [
-          {
-            role: "user",
-            content: [
-              { type: "text", text: "Describe this image." },
-              { type: "image", data: "fixture-image", mimeType: "image/png" },
-            ],
-            timestamp: 1,
-          },
+  it.each(["deepseek-flash", "deepseek-v4-flash-vision-exp"])(
+    "keeps image input in requests for %s",
+    (modelId) => {
+      const provider = buildDeepSeekProvider();
+      const entry = provider.models?.find((model) => model.id === modelId);
+      expect(entry).toMatchObject({ reasoning: true, input: ["text", "image"] });
+      const model = {
+        ...entry,
+        provider: "deepseek",
+        baseUrl: provider.baseUrl,
+        api: "openai-completions",
+      } as OpenAICompletionsModel;
+      const payload = buildOpenAICompletionsParams(
+        model,
+        {
+          messages: [
+            {
+              role: "user",
+              content: [
+                { type: "text", text: "Describe this image." },
+                { type: "image", data: "fixture-image", mimeType: "image/png" },
+              ],
+              timestamp: 1,
+            },
+          ],
+        },
+        {},
+      );
+      expect(payload.messages).toContainEqual({
+        role: "user",
+        content: [
+          { type: "text", text: "Describe this image." },
+          { type: "image_url", image_url: { url: "data:image/png;base64,fixture-image" } },
         ],
-      },
-      {},
-    );
-    expect(payload.messages).toContainEqual({
-      role: "user",
-      content: [
-        { type: "text", text: "Describe this image." },
-        { type: "image_url", image_url: { url: "data:image/png;base64,fixture-image" } },
-      ],
-    });
-  });
+      });
+    },
+  );
 
   it("adds blank reasoning_content for replayed tool calls from non-DeepSeek turns", async () => {
     const capture: PayloadCapture = {};

extensions/deepseek/models.ts

@@ -12,7 +12,9 @@ export const DEEPSEEK_MODEL_CATALOG: ModelDefinitionConfig[] = buildManifestMode
 }).models.map((model) => Object.assign(model, { api: "openai-completions" }));
 
 const DEEPSEEK_V4_MODEL_IDS = new Set(
-  DEEPSEEK_MODEL_CATALOG.map((model) => model.id).filter((id) => id.startsWith("deepseek-v4-")),
+  DEEPSEEK_MODEL_CATALOG.map((model) => model.id).filter(
+    (id) => id === "deepseek-flash" || id.startsWith("deepseek-v4-"),
+  ),
 );
 
 export function isDeepSeekV4ModelId(modelId: string): boolean {

extensions/deepseek/openclaw.plugin.json

@@ -92,6 +92,26 @@
               "maxTokensField": "max_tokens",
               "codeMode": "preferred"
             }
+          },
+          {
+            "id": "deepseek-flash",
+            "name": "DeepSeek V4.1 Flash",
+            "reasoning": true,
+            "input": ["text", "image"],
+            "contextWindow": 1000000,
+            "maxTokens": 384000,
+            "cost": {
+              "input": 0.3,
+              "output": 1.2,
+              "cacheRead": 0.006,
+              "cacheWrite": 0
+            },
+            "compat": {
+              "supportsUsageInStreaming": true,
+              "supportsReasoningEffort": true,
+              "maxTokensField": "max_tokens",
+              "codeMode": "preferred"
+            }
           }
         ]
       }