diff --git a/CHANGELOG.md b/CHANGELOG.md
index 081c8f7187b..0c6453204a9 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -59,7 +59,7 @@ Docs: https://docs.openclaw.ai
 
 ### Fixes
 
-- Gateway/startup: include the resolved thinking, reasoning, and fast-mode defaults in the `agent model` startup log line so operator logs show which mode new sessions will inherit.
+- Gateway/startup: include resolved thinking and fast-mode defaults in the `agent model` startup log line, defaulting unset startup thinking to `medium` without mixing in reasoning visibility.
 - Gateway/watch: suppress sync-I/O trace output during `pnpm gateway:watch --benchmark` unless explicitly requested, so CPU profiling no longer floods the terminal with stack traces.
 - Gateway/watch: when benchmark sync-I/O tracing is explicitly enabled, tee trace blocks to the benchmark output log and filter them from the terminal pane while keeping normal Gateway logs visible.
 - Agents/OpenAI: default direct OpenAI Responses models to the SSE transport instead of WebSocket auto-selection, preventing pi runtime chat turns from hanging on servers where the WebSocket path stalls while the OpenAI HTTP stream works. Thanks @vincentkoc.
diff --git a/docs/gateway/logging.md b/docs/gateway/logging.md
index 27624176054..c70a5b7252b 100644
--- a/docs/gateway/logging.md
+++ b/docs/gateway/logging.md
@@ -19,12 +19,12 @@ At startup, the Gateway logs the resolved default agent model together with the
 mode defaults that affect new sessions, for example:
 
 ```text
-agent model: openai-codex/gpt-5.5 (thinking=medium, reasoning=off, fast=on)
+agent model: openai-codex/gpt-5.5 (thinking=medium, fast=on)
 ```
 
 `thinking` comes from the default agent, model params, or global agent default;
-`reasoning` comes from the default agent or global reasoning default; and `fast`
-comes from the default agent or model `fastMode` params.
+when it is unset, the startup summary shows `medium`. `fast` comes from the
+default agent or model `fastMode` params.
 
 ## File-based logger
 
diff --git a/src/gateway/server-startup-log.test.ts b/src/gateway/server-startup-log.test.ts
index b2db0749ef1..781d0e3f3e0 100644
--- a/src/gateway/server-startup-log.test.ts
+++ b/src/gateway/server-startup-log.test.ts
@@ -49,7 +49,7 @@ describe("gateway startup log", () => {
     expect(warn).not.toHaveBeenCalled();
   });
 
-  it("logs configured model mode defaults with the startup model", () => {
+  it("logs configured model thinking and fast mode defaults with the startup model", () => {
     const info = vi.fn();
     const warn = vi.fn();
 
@@ -78,15 +78,50 @@ describe("gateway startup log", () => {
     });
 
     expect(info).toHaveBeenCalledWith(
-      "agent model: openai-codex/gpt-5.5 (thinking=medium, reasoning=stream, fast=on)",
+      "agent model: openai-codex/gpt-5.5 (thinking=medium, fast=on)",
       expect.objectContaining({
         consoleMessage: expect.stringContaining(
-          "agent model: openai-codex/gpt-5.5 (thinking=medium, reasoning=stream, fast=on)",
+          "agent model: openai-codex/gpt-5.5 (thinking=medium, fast=on)",
         ),
       }),
     );
   });
 
+  it("defaults unset startup thinking to medium", () => {
+    expect(
+      formatAgentModelStartupDetails({
+        cfg: {
+          agents: {
+            defaults: {
+              model: "openai-codex/gpt-5.5",
+            },
+            list: [{ id: "main", default: true, fastModeDefault: true }],
+          },
+        },
+        provider: "openai-codex",
+        model: "gpt-5.5",
+      }),
+    ).toBe("thinking=medium, fast=on");
+  });
+
+  it("preserves explicit startup thinking off", () => {
+    expect(
+      formatAgentModelStartupDetails({
+        cfg: {
+          agents: {
+            defaults: {
+              models: {
+                "openai-codex/gpt-5.5": { params: { thinking: "off", fastMode: true } },
+              },
+            },
+          },
+        },
+        provider: "openai-codex",
+        model: "gpt-5.5",
+      }),
+    ).toBe("thinking=off, fast=on");
+  });
+
   it("uses default agent mode overrides in the startup model details", () => {
     expect(
       formatAgentModelStartupDetails({
@@ -105,7 +140,7 @@ describe("gateway startup log", () => {
         provider: "openai",
         model: "gpt-5.5",
       }),
-    ).toBe("thinking=high, reasoning=off, fast=on");
+    ).toBe("thinking=high, fast=on");
   });
 
   it("logs a compact listening line with loaded plugin ids and duration", () => {
diff --git a/src/gateway/server-startup-log.ts b/src/gateway/server-startup-log.ts
index b50fdb48642..1e355072394 100644
--- a/src/gateway/server-startup-log.ts
+++ b/src/gateway/server-startup-log.ts
@@ -4,13 +4,24 @@ import { DEFAULT_MODEL, DEFAULT_PROVIDER } from "../agents/defaults.js";
 import { resolveFastModeState } from "../agents/fast-mode.js";
 import {
   resolveConfiguredModelRef,
-  resolveReasoningDefault,
   resolveThinkingDefault,
+  legacyModelKey,
+  modelKey,
 } from "../agents/model-selection.js";
 import type { OpenClawConfig } from "../config/types.openclaw.js";
 import { getResolvedLoggerSettings } from "../logging.js";
 import { collectEnabledInsecureOrDangerousFlags } from "../security/dangerous-config-flags.js";
 
+type StartupThinkLevel =
+  | "off"
+  | "minimal"
+  | "low"
+  | "medium"
+  | "high"
+  | "xhigh"
+  | "adaptive"
+  | "max";
+
 export function logGatewayStartup(params: {
   cfg: OpenClawConfig;
   bindHost: string;
@@ -57,6 +68,36 @@ export function logGatewayStartup(params: {
   }
 }
 
+function normalizeStartupThinkLevel(value: unknown): StartupThinkLevel | undefined {
+  return value === "off" ||
+    value === "minimal" ||
+    value === "low" ||
+    value === "medium" ||
+    value === "high" ||
+    value === "xhigh" ||
+    value === "adaptive" ||
+    value === "max"
+    ? value
+    : undefined;
+}
+
+function resolveExplicitStartupThinking(params: {
+  cfg: OpenClawConfig;
+  provider: string;
+  model: string;
+  defaultAgentThinking: unknown;
+}): StartupThinkLevel | undefined {
+  const models = params.cfg.agents?.defaults?.models;
+  const canonicalKey = modelKey(params.provider, params.model);
+  const legacyKey = legacyModelKey(params.provider, params.model);
+  return (
+    normalizeStartupThinkLevel(params.defaultAgentThinking) ??
+    normalizeStartupThinkLevel(models?.[canonicalKey]?.params?.thinking) ??
+    normalizeStartupThinkLevel(legacyKey ? models?.[legacyKey]?.params?.thinking : undefined) ??
+    normalizeStartupThinkLevel(params.cfg.agents?.defaults?.thinkingDefault)
+  );
+}
+
 export function formatAgentModelStartupDetails(params: {
   cfg: OpenClawConfig;
   provider: string;
@@ -64,20 +105,20 @@ export function formatAgentModelStartupDetails(params: {
 }): string {
   const defaultAgentId = resolveDefaultAgentId(params.cfg);
   const defaultAgentConfig = resolveAgentConfig(params.cfg, defaultAgentId);
-  const thinking =
-    defaultAgentConfig?.thinkingDefault ??
+  const explicitThinking = resolveExplicitStartupThinking({
+    cfg: params.cfg,
+    provider: params.provider,
+    model: params.model,
+    defaultAgentThinking: defaultAgentConfig?.thinkingDefault,
+  });
+  const resolvedThinking =
+    explicitThinking ??
     resolveThinkingDefault({
       cfg: params.cfg,
       provider: params.provider,
       model: params.model,
     });
-  const reasoning =
-    defaultAgentConfig?.reasoningDefault ??
-    params.cfg.agents?.defaults?.reasoningDefault ??
-    resolveReasoningDefault({
-      provider: params.provider,
-      model: params.model,
-    });
+  const thinking = explicitThinking ?? (resolvedThinking === "off" ? "medium" : resolvedThinking);
   const fast = resolveFastModeState({
     cfg: params.cfg,
     provider: params.provider,
@@ -85,7 +126,7 @@ export function formatAgentModelStartupDetails(params: {
     agentId: defaultAgentId,
   });
 
-  return `thinking=${thinking}, reasoning=${reasoning}, fast=${fast.enabled ? "on" : "off"}`;
+  return `thinking=${thinking}, fast=${fast.enabled ? "on" : "off"}`;
 }
 
 function formatReadyDetails(