diff --git a/docs-site/src/content/docs/guides/grok-build.md b/docs-site/src/content/docs/guides/grok-build.md
index 958b25a30b..75dfadf84e 100644
--- a/docs-site/src/content/docs/guides/grok-build.md
+++ b/docs-site/src/content/docs/guides/grok-build.md
@@ -61,6 +61,13 @@ in Codex. Models with an empty tier list keep no effort control, matching Codex
 behavior. Native GPT-5.6 entries are separate: they preserve and expose their pinned
 upstream reasoning ladders rather than provider-configured routed metadata.
 
+Grok Build talks to opencodex over Chat Completions and sends `reasoning_effort` when
+the ladder is advertised. The Chat Completions inbound translator defaults the internal
+Responses `reasoning.summary` to `auto` in that case, so thinking traces reach Grok as
+`delta.reasoning_content` instead of being hidden. Set `include_reasoning: false` (or
+`reasoning.summary: "none"`) if a client wants the model to think without returning the
+trace. An explicit `reasoning.summary` wins when both knobs are present.
+
 ## Authentication note
 
 Grok Build requires a non-empty API key for custom models even on loopback. The injected
diff --git a/docs-site/src/content/docs/reference/proxy-formats.md b/docs-site/src/content/docs/reference/proxy-formats.md
index bd02453019..72bc565121 100644
--- a/docs-site/src/content/docs/reference/proxy-formats.md
+++ b/docs-site/src/content/docs/reference/proxy-formats.md
@@ -145,6 +145,14 @@ non-empty `messages` array. It translates system, user, assistant, and tool mess
 Responses items; translates function tools, tool choice, images, reasoning effort, and supported
 response formats; runs the normal Responses routing pipeline; then translates the result back.
 
+Reasoning is part of that translation. `reasoning_effort` (or `reasoning.effort`) becomes
+internal `reasoning.effort`. Because the Responses parser hides thinking unless
+`reasoning.summary` is set and is not `none`, Chat Completions requests that ask for an
+effort default to `reasoning.summary: "auto"` so thinking streams back as
+`delta.reasoning_content`. Clients can still hide traces with `include_reasoning: false` or
+`reasoning.summary: "none"`. An explicit `reasoning.summary` of `auto`, `concise`,
+`detailed`, or `none` wins over `include_reasoning`.
+
 Structured output is part of that translation: `response_format` with `json_object` or
 `json_schema` is forwarded to routed `openai-chat` models. On `POST /v1/responses` the
 equivalent request field is `text.format`: native Responses routes preserve it in the raw
diff --git a/src/chat/inbound.ts b/src/chat/inbound.ts
index f9abe383ad..ff3dba5e4b 100644
--- a/src/chat/inbound.ts
+++ b/src/chat/inbound.ts
@@ -14,6 +14,7 @@ function isRec(v: unknown): v is Rec {
 }
 
 const OUTPUT_CONFIG_EFFORTS = new Set(["minimal", "low", "medium", "high", "xhigh", "max", "ultra"]);
+const OUTPUT_CONFIG_SUMMARIES = new Set(["auto", "concise", "detailed", "none"]);
 
 function contentToText(content: unknown): string {
   if (typeof content === "string") return content;
@@ -205,6 +206,22 @@ function resolveReasoningEffort(raw: Rec): string | undefined {
   return undefined;
 }
 
+/**
+ * Chat Completions clients (Grok Build, Copilot, OpenAI-compatible SDKs) expect
+ * `delta.reasoning_content` whenever the model thinks. The internal Responses
+ * parser hides thinking unless `reasoning.summary` is set and is not `"none"`.
+ * Map the common Chat Completions knobs onto that field. When the client only
+ * sent an effort, default the summary to `"auto"` so traces are not swallowed.
+ */
+function resolveReasoningSummary(raw: Rec): string | undefined {
+  if (isRec(raw.reasoning) && typeof raw.reasoning.summary === "string" && OUTPUT_CONFIG_SUMMARIES.has(raw.reasoning.summary)) {
+    return raw.reasoning.summary;
+  }
+  if (raw.include_reasoning === false) return "none";
+  if (raw.include_reasoning === true) return "auto";
+  return undefined;
+}
+
 /**
  * Translate an OpenAI Chat Completions request body into a /v1/responses request body.
  * Throws ChatCompletionsRequestError (-> 400) on malformed input.
@@ -286,7 +303,13 @@ export function chatCompletionsToResponsesBody(raw: unknown): Rec {
   if (raw.metadata !== undefined) body.metadata = raw.metadata;
 
   const effort = resolveReasoningEffort(raw);
-  if (effort) body.reasoning = { effort };
+  const summary = resolveReasoningSummary(raw);
+  if (effort || summary !== undefined) {
+    body.reasoning = {
+      ...(effort ? { effort } : {}),
+      summary: summary ?? "auto",
+    };
+  }
 
   const text = responseFormatToText(raw.response_format);
   if (text) body.text = text;
diff --git a/src/server/chat-completions.ts b/src/server/chat-completions.ts
index 34864ad834..1ed1d4e6d6 100644
--- a/src/server/chat-completions.ts
+++ b/src/server/chat-completions.ts
@@ -140,9 +140,11 @@ async function handleChatCompletionsWithBudget(
       logCtx.usageLogInputTokens = Math.max(1, estimateTokens(parts.join("\n"), requestedModel));
     }
     if (internalBody.reasoning !== undefined) {
-      const { supportedLadderFor } = await import("./effort-policy");
+      const { stripEmptyLadderEffort, supportedLadderFor } = await import("./effort-policy");
       const ladder = supportedLadderFor({ provider: route.provider, modelId: route.modelId });
-      if (ladder !== undefined && ladder.length === 0) delete internalBody.reasoning;
+      const next = stripEmptyLadderEffort(internalBody.reasoning, ladder);
+      if (next === undefined) delete internalBody.reasoning;
+      else internalBody.reasoning = next;
     }
   } catch (err) {
     if (err instanceof NoEligiblePolicyCandidateError) {
diff --git a/src/server/effort-policy.ts b/src/server/effort-policy.ts
index 57f8b99e00..2686b73460 100644
--- a/src/server/effort-policy.ts
+++ b/src/server/effort-policy.ts
@@ -98,6 +98,24 @@ export function effortCapAppliesTo(
  * validates against that backend. A custom responses provider (key mode) serving a
  * native-looking bare id must NOT inherit the unrelated native ladder.
  */
+/**
+ * Empty ladders mean "no effort control", not "this model cannot emit reasoning".
+ * Drop only `effort` so a Chat Completions `include_reasoning` / `reasoning.summary`
+ * request still reaches parseRequest and is not hidden by hideThinkingSummary.
+ */
+export function stripEmptyLadderEffort(
+  reasoning: unknown,
+  ladder: readonly string[] | undefined,
+): unknown {
+  if (ladder === undefined || ladder.length > 0) return reasoning;
+  if (reasoning === undefined || reasoning === null || typeof reasoning !== "object" || Array.isArray(reasoning)) {
+    return reasoning;
+  }
+  const next = { ...(reasoning as Record<string, unknown>) };
+  delete next.effort;
+  return Object.keys(next).length > 0 ? next : undefined;
+}
+
 export function supportedLadderFor(route: { provider: OcxProviderConfig; modelId: string }): string[] | undefined {
   const { provider, modelId } = route;
   if (modelInList(provider.noReasoningModels, modelId)) return [];
diff --git a/tests/chat-completions-endpoint.test.ts b/tests/chat-completions-endpoint.test.ts
index eea0719eb3..a7ecca3993 100644
--- a/tests/chat-completions-endpoint.test.ts
+++ b/tests/chat-completions-endpoint.test.ts
@@ -1,5 +1,5 @@
 import { waitForNativeMainStartupGate } from "../src/codex/native-profile-startup";
-import { afterEach, beforeEach, expect, test } from "bun:test";
+import { afterEach, beforeEach, describe, expect, test } from "bun:test";
 import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
@@ -10,6 +10,7 @@ import type { OcxConfig, OcxProviderConfig } from "../src/types";
 import { installIsolatedCodexHome, type IsolatedCodexHome } from "./helpers/isolated-codex-home";
 import { chatCompletionsToResponsesBody, ChatCompletionsRequestError } from "../src/chat/inbound";
 import { chatCompletionsUsage } from "../src/chat/outbound";
+import { parseRequest } from "../src/responses/parser";
 import { createTestTranslatorBudget } from "./helpers/translator-budget";
 import type { TranslatorBudget } from "../src/lib/translator-budget";
 import {
@@ -218,7 +219,7 @@ test("chatCompletionsToResponsesBody maps messages/tools/system", () => {
   expect(body.stream).toBe(true);
   expect(body.instructions).toBe("be brief");
   expect(body.max_output_tokens).toBe(64);
-  expect(body.reasoning).toEqual({ effort: "high" });
+  expect(body.reasoning).toEqual({ effort: "high", summary: "auto" });
   expect(body.tool_choice).toBe("auto");
   expect(Array.isArray(body.tools)).toBe(true);
   expect((body.tools as Array<Record<string, unknown>>)[0]).toMatchObject({ type: "function", name: "lookup" });
@@ -228,6 +229,81 @@ test("chatCompletionsToResponsesBody maps messages/tools/system", () => {
   expect(input.some(i => i.type === "function_call_output" && i.call_id === "call_1")).toBe(true);
 });
 
+describe("chatCompletionsToResponsesBody reasoning summary", () => {
+  test("defaults summary to auto when the client only sent reasoning_effort", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "opencode-go/deepseek-v4-flash",
+      messages: [{ role: "user", content: "What is 17*19?" }],
+      reasoning_effort: "max",
+    });
+    expect(body.reasoning).toEqual({ effort: "max", summary: "auto" });
+    const parsed = parseRequest(body);
+    expect(parsed.options.hideThinkingSummary).not.toBe(true);
+    expect(parsed.options.reasoning).toBe("max");
+  });
+
+  test("preserves an explicit reasoning.summary", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+      reasoning: { effort: "high", summary: "concise" },
+    });
+    expect(body.reasoning).toEqual({ effort: "high", summary: "concise" });
+  });
+
+  test("include_reasoning false hides thinking even when effort is set", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+      reasoning_effort: "high",
+      include_reasoning: false,
+    });
+    expect(body.reasoning).toEqual({ effort: "high", summary: "none" });
+    expect(parseRequest(body).options.hideThinkingSummary).toBe(true);
+  });
+
+  test("include_reasoning true requests a visible summary without an effort", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+      include_reasoning: true,
+    });
+    expect(body.reasoning).toEqual({ summary: "auto" });
+    expect(parseRequest(body).options.hideThinkingSummary).not.toBe(true);
+  });
+
+  test("explicit reasoning.summary wins over include_reasoning true", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+      include_reasoning: true,
+      reasoning: { summary: "none" },
+    });
+    expect(body.reasoning).toEqual({ summary: "none" });
+    expect(parseRequest(body).options.hideThinkingSummary).toBe(true);
+  });
+
+  test("explicit reasoning.summary wins over include_reasoning false", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+      include_reasoning: false,
+      reasoning: { effort: "high", summary: "auto" },
+    });
+    expect(body.reasoning).toEqual({ effort: "high", summary: "auto" });
+    expect(parseRequest(body).options.hideThinkingSummary).not.toBe(true);
+  });
+
+  test("omits reasoning when the client sent no reasoning knobs", () => {
+    const body = chatCompletionsToResponsesBody({
+      model: "mock/test-model",
+      messages: [{ role: "user", content: "hi" }],
+    });
+    expect(body.reasoning).toBeUndefined();
+    expect(parseRequest(body).options.hideThinkingSummary).toBe(true);
+  });
+});
+
 test("chatCompletionsToResponsesBody rejects missing model", () => {
   expect(() => chatCompletionsToResponsesBody({ messages: [{ role: "user", content: "x" }] }))
     .toThrow(ChatCompletionsRequestError);
diff --git a/tests/effort-policy.test.ts b/tests/effort-policy.test.ts
index 1c26b40dd8..4cd88d8039 100644
--- a/tests/effort-policy.test.ts
+++ b/tests/effort-policy.test.ts
@@ -7,7 +7,7 @@ import { afterEach, describe, expect, test } from "bun:test";
 import { mkdtempSync, rmSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
-import { applyEffortCap, effortCapAppliesTo, effortCapFor, isThreadSpawnRequest, resolveCappedEffort, supportedLadderFor } from "../src/server/effort-policy";
+import { applyEffortCap, effortCapAppliesTo, effortCapFor, isThreadSpawnRequest, resolveCappedEffort, stripEmptyLadderEffort, supportedLadderFor } from "../src/server/effort-policy";
 import { collabSurface } from "../src/server/responses";
 import { handleManagementAPI } from "../src/server/management-api";
 import { NoEnabledOpenAiProviderError, routeModel } from "../src/router";
@@ -173,6 +173,26 @@ describe("resolveCappedEffort (ladder-aware resolution)", () => {
   });
 });
 
+describe("stripEmptyLadderEffort", () => {
+  test("keeps a summary-only object on an empty ladder", () => {
+    expect(stripEmptyLadderEffort({ summary: "auto" }, [])).toEqual({ summary: "auto" });
+  });
+
+  test("strips effort but keeps summary on an empty ladder", () => {
+    expect(stripEmptyLadderEffort({ effort: "max", summary: "auto" }, [])).toEqual({ summary: "auto" });
+  });
+
+  test("drops an effort-only object on an empty ladder", () => {
+    expect(stripEmptyLadderEffort({ effort: "high" }, [])).toBeUndefined();
+  });
+
+  test("leaves reasoning untouched when the ladder is unknown or non-empty", () => {
+    const reasoning = { effort: "high", summary: "auto" };
+    expect(stripEmptyLadderEffort(reasoning, undefined)).toBe(reasoning);
+    expect(stripEmptyLadderEffort(reasoning, ["high"])).toBe(reasoning);
+  });
+});
+
 describe("applyEffortCap strip paths", () => {
   test("no-effort model strips even a below-cap effort from BOTH shapes, keeping summary", () => {
     const config = makeConfig({ effortCap: "high" });