diff --git a/extensions/ollama/src/setup-model-selection.test.ts b/extensions/ollama/src/setup-model-selection.test.ts
index 5868a5528135..9e2abb615a78 100644
--- a/extensions/ollama/src/setup-model-selection.test.ts
+++ b/extensions/ollama/src/setup-model-selection.test.ts
@@ -1,3 +1,4 @@
+import type { WizardPrompter } from "openclaw/plugin-sdk/setup";
 import { requestUrl } from "openclaw/plugin-sdk/test-env";
 import { afterEach, describe, expect, it, vi } from "vitest";
 import {
@@ -8,6 +9,7 @@ import {
   normalizeOllamaModelName,
   selectAppGuidedOllamaModelFromDiscovery,
 } from "./setup-model-selection.js";
+import { promptAndConfigureOllama } from "./setup.js";
 
 afterEach(() => {
   vi.unstubAllGlobals();
@@ -81,13 +83,24 @@ describe("Ollama onboarding model selection", () => {
     ).toBe("qwen3:0.6b");
   });
 
-  it("prefers the smallest non-reasoning setup model", () => {
+  it.each([
+    {
+      description: "skips a smaller model with an explicit thinking capability",
+      capabilities: ["tools", "thinking"],
+      expected: "llama3.2:latest",
+    },
+    {
+      description: "trusts explicit non-thinking capabilities over a reasoning-like model name",
+      capabilities: ["tools"],
+      expected: "deepseek-r1:8b",
+    },
+  ])("$description", ({ capabilities, expected }) => {
     expect(
       selectAppGuidedOllamaModelFromDiscovery([
         {
           name: "deepseek-r1:8b",
           contextWindow: 131_072,
-          capabilities: ["tools", "thinking"],
+          capabilities,
           size: 1_000,
         },
         {
@@ -103,7 +116,44 @@ describe("Ollama onboarding model selection", () => {
           size: 2_000,
         },
       ]),
-    ).toBe("llama3.2:latest");
+    ).toBe(expected);
+  });
+
+  it("selects the smallest explicitly non-thinking model during interactive setup", async () => {
+    vi.stubGlobal(
+      "fetch",
+      vi.fn(async (input: string | URL | Request) => {
+        const url = requestUrl(input);
+        if (url.endsWith("/api/tags")) {
+          return Response.json({
+            models: [
+              { name: "deepseek-r1:1b", size: 100 },
+              { name: "llama3:70b", size: 70_000 },
+            ],
+          });
+        }
+        if (url.endsWith("/api/show")) {
+          return Response.json({
+            model_info: { "test.context_length": 32_768 },
+            capabilities: ["completion", "tools"],
+          });
+        }
+        throw new Error(`Unexpected fetch: ${url}`);
+      }),
+    );
+
+    const prompter = {
+      select: vi.fn().mockResolvedValueOnce("local-only"),
+      text: vi.fn().mockResolvedValueOnce("http://127.0.0.1:11434"),
+      note: vi.fn(async () => undefined),
+    } as unknown as WizardPrompter;
+    const result = await promptAndConfigureOllama({ cfg: {}, prompter });
+    const selected = result.config.models?.providers?.ollama?.models?.find(
+      (model) => model.id === "deepseek-r1:1b",
+    );
+
+    expect(selected?.reasoning).toBe(false);
+    expect(result.defaultModel).toBe("ollama/deepseek-r1:1b");
   });
 
   it("aborts pending model discovery with the setup signal", async () => {
diff --git a/extensions/ollama/src/setup-model-selection.ts b/extensions/ollama/src/setup-model-selection.ts
index 565faf2cf554..b833932d3041 100644
--- a/extensions/ollama/src/setup-model-selection.ts
+++ b/extensions/ollama/src/setup-model-selection.ts
@@ -112,8 +112,7 @@ export function selectAppGuidedOllamaModelFromDiscovery(
       id: model.name,
       contextWindow: model.contextWindow,
       supportsTools: model.capabilities?.includes("tools") === true,
-      reasoning:
-        model.capabilities?.includes("thinking") === true || isReasoningModelHeuristic(model.name),
+      reasoning: model.capabilities?.includes("thinking") ?? isReasoningModelHeuristic(model.name),
       size: model.size,
     })),
   );