diff --git a/open-sse/config/kiroConstants.js b/open-sse/config/kiroConstants.js
index 93b8f4569e..511ae91935 100644
--- a/open-sse/config/kiroConstants.js
+++ b/open-sse/config/kiroConstants.js
@@ -8,8 +8,8 @@
  *   - `-agentic` model suffix detection + chunked-write system prompt
  *   - reasoning / thinking trigger detection (Anthropic-Beta header,
  *     Claude `thinking`, OpenAI `reasoning_effort`, AMP/Cursor magic tag)
- *   - the `<thinking_mode>enabled</thinking_mode>` system-prompt injection
- *     that turns Kiro reasoning on
+ *   - schema-specific native effort fields for supported GPT and Claude models
+ *   - legacy `<thinking_mode>` system-prompt injection for other models
  *
  * Kiro upstream does not advertise `-agentic` model IDs; they are a 9router
  * fiction. The suffix is stripped before the request leaves this process.
@@ -109,6 +109,7 @@ export function resolveKiroThinkingBudget(body, headers, model) {
   const cfg = extractThinking(body);
   if (cfg) {
     if (cfg.mode === "none") return null;
+    if (cfg.mode === "level" && cfg.level === "disabled") return null;
     if (cfg.mode === "budget") return cfg.budget;
     if (cfg.mode === "level") return effortToBudget(cfg.level) ?? KIRO_THINKING_BUDGET_DEFAULT;
     return KIRO_THINKING_BUDGET_DEFAULT;
@@ -144,9 +145,30 @@ export function extractKiroEffortLevel(body) {
   return null;
 }
 
-export function buildKiroAdditionalModelRequestFields(body) {
-  const effort = extractKiroEffortLevel(body);
+function extractKiroGptEffortLevel(body) {
+  const effort =
+    body?.output_config?.effort ??
+    body?.reasoning_effort ??
+    (typeof body?.reasoning === "object" ? body.reasoning?.effort : null);
+  if (typeof effort !== "string") return null;
+  const normalized = effort.toLowerCase();
+  if (normalized === "max") return "xhigh";
+  // Kiro CLI does not advertise an explicit GPT "none" wire value; omit it.
+  if (["low", "medium", "high", "xhigh"].includes(normalized)) {
+    return normalized;
+  }
+  return null;
+}
+
+export function buildKiroAdditionalModelRequestFields(body, effortPath = "output_config") {
+  const effort = effortPath === "reasoning"
+    ? extractKiroGptEffortLevel(body)
+    : extractKiroEffortLevel(body);
   if (!effort) return undefined;
+  if (effortPath === "reasoning") {
+    // Mirrors Kiro CLI/KAS buildEffortRequestFields("reasoning") for GPT.
+    return { reasoning: { effort } };
+  }
   // Mirrors Kiro CLI/KAS buildEffortRequestFields("output_config").
   return {
     thinking: { type: "adaptive", display: "summarized" },
@@ -154,12 +176,15 @@ export function buildKiroAdditionalModelRequestFields(body) {
   };
 }
 
-export function supportsKiroAdditionalModelRequestFields(model) {
-  if (typeof model !== "string") return false;
+export function resolveKiroEffortPath(model) {
+  if (typeof model !== "string") return null;
   const normalized = model.toLowerCase().replace(/-/g, ".");
-  if (!normalized.includes("claude")) return false;
+  if (/(?:^|[/.])gpt[/.]5[/.]6(?:[/.]|$)/.test(normalized)) {
+    return "reasoning";
+  }
+  if (!normalized.includes("claude")) return null;
   const match = normalized.match(/(?:^|[/.])claude(?:[/.][a-z]+)*[/.](\d+)(?:[/.](\d+))?(?:[/.]|$)/);
-  if (!match) return false;
+  if (!match) return null;
   const [, majorText, minorText] = match;
   const major = Number(majorText);
   const minor = minorText === undefined ? null : Number(minorText);
@@ -167,12 +192,24 @@ export function supportsKiroAdditionalModelRequestFields(model) {
   // Kiro rejected additionalModelRequestFields on legacy 4.5 models in live smoke.
   // Default future Claude/Kiro models to supported so new model releases do not
   // need a code allowlist update.
-  return !(major < 4 || (major === 4 && (minor === null || minor <= 5 || dateSuffixMinor)));
+  return major < 4 || (major === 4 && (minor === null || minor <= 5 || dateSuffixMinor))
+    ? null
+    : "output_config";
+}
+
+export function supportsKiroAdditionalModelRequestFields(model) {
+  return resolveKiroEffortPath(model) !== null;
+}
+
+export function usesKiroNativeGptEffort(body, model) {
+  return resolveKiroEffortPath(model) === "reasoning"
+    && extractKiroGptEffortLevel(body) !== null;
 }
 
 export function buildKiroAdditionalModelRequestFieldsForModel(body, model) {
-  if (!supportsKiroAdditionalModelRequestFields(model)) return undefined;
-  return buildKiroAdditionalModelRequestFields(body);
+  const effortPath = resolveKiroEffortPath(model);
+  if (!effortPath) return undefined;
+  return buildKiroAdditionalModelRequestFields(body, effortPath);
 }
 
 /**
diff --git a/open-sse/translator/request/claude-to-kiro.js b/open-sse/translator/request/claude-to-kiro.js
index bcb8b97fe0..98531c736d 100644
--- a/open-sse/translator/request/claude-to-kiro.js
+++ b/open-sse/translator/request/claude-to-kiro.js
@@ -33,6 +33,7 @@ import {
   KIRO_AGENTIC_SYSTEM_PROMPT,
   resolveDefaultProfileArn,
   buildKiroAdditionalModelRequestFieldsForModel,
+  usesKiroNativeGptEffort,
 } from "../../config/kiroConstants.js";
 import { DEFAULT_IMAGE_MIME } from "../schema/index.js";
 import { ROLE, CLAUDE_BLOCK } from "../schema/index.js";
@@ -390,6 +391,8 @@ export function claudeToKiroRequest(model, body, stream, credentials) {
 
   const { upstream: upstreamModel, agentic } = resolveKiroModel(model);
   const thinkingBudget = resolveKiroThinkingBudget(body, credentials?.rawHeaders, model);
+  const additionalModelRequestFields = buildKiroAdditionalModelRequestFieldsForModel(body, upstreamModel);
+  const usesNativeGptEffort = usesKiroNativeGptEffort(body, upstreamModel);
 
   // Guard 1: no client tools → flatten all tool interactions to text.
   if (!clientProvidedTools) {
@@ -421,7 +424,9 @@ export function claudeToKiroRequest(model, body, stream, credentials) {
   // enforce top-level systemPrompt for direct calls.
   const timestamp = new Date().toISOString();
   const systemPromptParts = [];
-  if (thinkingBudget !== null) systemPromptParts.push(buildThinkingSystemPrefix(thinkingBudget));
+  if (thinkingBudget !== null && !usesNativeGptEffort) {
+    systemPromptParts.push(buildThinkingSystemPrefix(thinkingBudget));
+  }
   if (agentic) systemPromptParts.push(KIRO_AGENTIC_SYSTEM_PROMPT);
   const systemInstruction = extractClaudeSystemText(body.system);
   if (systemInstruction) systemPromptParts.push(systemInstruction);
@@ -481,7 +486,6 @@ export function claudeToKiroRequest(model, body, stream, credentials) {
 
   if (profileArn) payload.profileArn = profileArn;
   if (systemPrompt) payload.systemPrompt = systemPrompt;
-  const additionalModelRequestFields = buildKiroAdditionalModelRequestFieldsForModel(body, upstreamModel);
   if (additionalModelRequestFields) {
     payload.additionalModelRequestFields = additionalModelRequestFields;
   }
diff --git a/open-sse/translator/request/openai-responses.js b/open-sse/translator/request/openai-responses.js
index 1fc058250a..64740c12cf 100644
--- a/open-sse/translator/request/openai-responses.js
+++ b/open-sse/translator/request/openai-responses.js
@@ -200,6 +200,9 @@ export function openaiResponsesToOpenAIRequest(model, body, stream, credentials)
   delete result.include;
   delete result.prompt_cache_key;
   delete result.store;
+  if (typeof result.reasoning?.effort === "string") {
+    result.reasoning_effort = result.reasoning.effort;
+  }
   delete result.reasoning;
   delete result.client_metadata;
 
diff --git a/open-sse/translator/request/openai-to-kiro.js b/open-sse/translator/request/openai-to-kiro.js
index a2681b5a7d..069feffa9a 100644
--- a/open-sse/translator/request/openai-to-kiro.js
+++ b/open-sse/translator/request/openai-to-kiro.js
@@ -13,7 +13,8 @@ import {
   buildThinkingSystemPrefix,
   KIRO_AGENTIC_SYSTEM_PROMPT,
   resolveDefaultProfileArn,
-  buildKiroAdditionalModelRequestFieldsForModel
+  buildKiroAdditionalModelRequestFieldsForModel,
+  usesKiroNativeGptEffort
 } from "../../config/kiroConstants.js";
 import { parseDataUri } from "../concerns/image.js";
 import { DEFAULT_IMAGE_MIME } from "../schema/index.js";
@@ -511,12 +512,10 @@ function convertMessages(messages, tools, model) {
  *    Kiro's 2-3 minute server timeout. The suffix is stripped before being
  *    sent upstream.
  *
- * 2. Thinking / reasoning. Kiro does not accept `thinking.type` or
- *    `reasoning_effort` natively. The only way to enable reasoning is to
- *    inject `<thinking_mode>enabled</thinking_mode>` into the user content
- *    sent upstream. Detection covers Anthropic-Beta header, Claude API
+ * 2. Thinking / reasoning. Detection covers Anthropic-Beta header, Claude API
  *    `thinking`, OpenAI `reasoning_effort`, AMP/Cursor magic tags, and model
- *    name hints.
+ *    name hints. Supported models receive Kiro's schema-specific effort fields;
+ *    legacy prompt tags remain only for models that need them.
  */
 export function openaiToKiroRequest(model, body, stream, credentials) {
   const messages = body.messages || [];
@@ -527,6 +526,8 @@ export function openaiToKiroRequest(model, body, stream, credentials) {
 
   const { upstream: upstreamModel, agentic } = resolveKiroModel(model);
   const thinkingBudget = resolveKiroThinkingBudget(body, credentials?.rawHeaders, model);
+  const additionalModelRequestFields = buildKiroAdditionalModelRequestFieldsForModel(body, upstreamModel);
+  const usesNativeGptEffort = usesKiroNativeGptEffort(body, upstreamModel);
 
   const { history, currentMessage } = convertMessages(messages, tools, upstreamModel);
 
@@ -554,7 +555,7 @@ export function openaiToKiroRequest(model, body, stream, credentials) {
   // too because the CodeWhisperer surface does not always enforce top-level
   // systemPrompt for direct calls.
   const systemPromptParts = [];
-  if (thinkingBudget !== null) {
+  if (thinkingBudget !== null && !usesNativeGptEffort) {
     systemPromptParts.push(buildThinkingSystemPrefix(thinkingBudget));
   }
   if (agentic) {
@@ -612,7 +613,6 @@ export function openaiToKiroRequest(model, body, stream, credentials) {
     payload.profileArn = profileArn;
   }
   if (systemPrompt) payload.systemPrompt = systemPrompt;
-  const additionalModelRequestFields = buildKiroAdditionalModelRequestFieldsForModel(body, upstreamModel);
   if (additionalModelRequestFields) {
     payload.additionalModelRequestFields = additionalModelRequestFields;
   }
diff --git a/tests/translator/bugs-kiro.test.js b/tests/translator/bugs-kiro.test.js
index 7f3677facc..c3f777d475 100644
--- a/tests/translator/bugs-kiro.test.js
+++ b/tests/translator/bugs-kiro.test.js
@@ -5,8 +5,34 @@ import { translateRequest } from "../../open-sse/translator/index.js";
 import { FORMATS } from "../../open-sse/translator/formats.js";
 
 const O2K = (body) => translateRequest(FORMATS.OPENAI, FORMATS.KIRO, "m", body, true, null, "kiro");
+const R2K = (model, body) => translateRequest(
+  FORMATS.OPENAI_RESPONSES,
+  FORMATS.KIRO,
+  model,
+  body,
+  true,
+  null,
+  "kiro"
+);
 
 describe("OpenAI → Kiro", () => {
+  it.each([
+    ["high", "gpt-5.6-sol"],
+    ["medium", "gpt-5.6-terra"],
+    ["low", "gpt-5.6-luna"],
+  ])("preserves Responses reasoning.effort %s through the full Kiro route", (effort, model) => {
+    const out = R2K(model, {
+      input: "Use the requested effort",
+      reasoning: { effort },
+    });
+
+    expect(out.additionalModelRequestFields).toEqual({
+      reasoning: { effort },
+    });
+    expect(out.systemPrompt || "").not.toContain("<thinking_mode>");
+    expect(out.systemPrompt || "").not.toContain("<max_thinking_length>");
+  });
+
   // openai-to-kiro.js — safeJSONParse guards bad tool-call JSON (fixed in PR #1582)
   it("malformed tool arguments do not throw the whole request", () => {
     expect(() =>
diff --git a/tests/translator/claude-kiro-direct.test.js b/tests/translator/claude-kiro-direct.test.js
index 01e217092a..3c2d964b33 100644
--- a/tests/translator/claude-kiro-direct.test.js
+++ b/tests/translator/claude-kiro-direct.test.js
@@ -112,6 +112,59 @@ describe("Claude → Kiro (direct route)", () => {
     expect(out.systemPrompt).toContain("<max_thinking_length>24576</max_thinking_length>");
   });
 
+  it("maps Claude-format effort to GPT-5.6 reasoning fields without legacy prompt tags", () => {
+    const out = C2K({
+      output_config: { effort: "low" },
+      messages: [{ role: "user", content: "think lightly" }],
+    }, null, "gpt-5.6-sol");
+
+    expect(out.additionalModelRequestFields).toEqual({
+      reasoning: { effort: "low" },
+    });
+    expect(out.systemPrompt || "").not.toContain("<thinking_mode>");
+    expect(out.systemPrompt || "").not.toContain("<max_thinking_length>");
+  });
+
+  it.each(["auto", "minimal", "ultra"])(
+    "keeps the legacy thinking fallback for unsupported GPT-5.6 effort %s",
+    (effort) => {
+      const out = C2K({
+        output_config: { effort },
+        messages: [{ role: "user", content: "Use legacy thinking" }],
+      }, null, "gpt-5.6-sol");
+
+      expect(out.additionalModelRequestFields).toBeUndefined();
+      expect(out.systemPrompt).toContain("<thinking_mode>enabled</thinking_mode>");
+      expect(out.systemPrompt).toContain("<max_thinking_length>");
+    }
+  );
+
+  it.each(["none", "off", "disabled"])(
+    "keeps GPT-5.6 reasoning intentionally disabled for effort %s",
+    (effort) => {
+      const out = C2K({
+        output_config: { effort },
+        messages: [{ role: "user", content: "Do not reason" }],
+      }, null, "gpt-5.6-sol");
+
+      expect(out.additionalModelRequestFields).toBeUndefined();
+      expect(out.systemPrompt || "").not.toContain("<thinking_mode>");
+      expect(out.systemPrompt || "").not.toContain("<max_thinking_length>");
+    }
+  );
+
+  it("keeps explicit Claude effort ahead of an injected OpenAI effort", () => {
+    const out = C2K({
+      output_config: { effort: "low" },
+      reasoning_effort: "high",
+      messages: [{ role: "user", content: "honor the client effort" }],
+    }, null, "gpt-5.6-sol");
+
+    expect(out.additionalModelRequestFields).toEqual({
+      reasoning: { effort: "low" },
+    });
+  });
+
   it("sends Claude system as top-level systemPrompt and keeps a user-content fallback", () => {
     const out = C2K({
       system: "system-only instruction",
diff --git a/tests/unit/openai-to-kiro.test.js b/tests/unit/openai-to-kiro.test.js
index 8b1c97e868..dbaf7c85c7 100644
--- a/tests/unit/openai-to-kiro.test.js
+++ b/tests/unit/openai-to-kiro.test.js
@@ -316,6 +316,102 @@ describe("openaiToKiroRequest", () => {
       });
     });
 
+    it.each([
+      ["high", "gpt-5.6-sol"],
+      ["medium", "kiro/gpt-5.6-terra"],
+      ["low", "gpt-5.6-luna"],
+    ])("maps GPT-5.6 reasoning.effort %s without legacy prompt tags", (effort, model) => {
+      const body = {
+        reasoning: { effort },
+        messages: [{ role: "user", content: "Use the requested effort" }]
+      };
+
+      const result = openaiToKiroRequest(model, body, true, {});
+
+      expect(result.additionalModelRequestFields).toEqual({
+        reasoning: { effort },
+      });
+      expect(systemPromptOf(result)).not.toContain("<thinking_mode>");
+      expect(systemPromptOf(result)).not.toContain("<max_thinking_length>");
+      expect(contentOf(result)).not.toContain("<thinking_mode>");
+      expect(contentOf(result)).not.toContain("<max_thinking_length>");
+    });
+
+    it.each([
+      ["xhigh", "gpt-5.6-terra", "xhigh"],
+      ["max", "gpt-5.6-sol", "xhigh"],
+    ])("preserves GPT-5.6 effort %s as supported wire effort %s", (effort, model, wireEffort) => {
+      const body = {
+        reasoning: { effort },
+        messages: [{ role: "user", content: "Use extended effort" }]
+      };
+
+      const result = openaiToKiroRequest(model, body, true, {});
+
+      expect(result.additionalModelRequestFields).toEqual({
+        reasoning: { effort: wireEffort },
+      });
+      expect(systemPromptOf(result)).not.toContain("<thinking_mode>");
+      expect(systemPromptOf(result)).not.toContain("<max_thinking_length>");
+    });
+
+    it("omits GPT-5.6 effort fields and legacy prompt tags when effort is absent", () => {
+      const body = {
+        messages: [{ role: "user", content: "No explicit reasoning effort" }]
+      };
+
+      const result = openaiToKiroRequest("gpt-5.6-sol", body, true, {});
+
+      expect(result.additionalModelRequestFields).toBeUndefined();
+      expect(systemPromptOf(result)).not.toContain("<thinking_mode>");
+      expect(systemPromptOf(result)).not.toContain("<max_thinking_length>");
+    });
+
+    it.each(["auto", "minimal", "ultra"])(
+      "keeps the legacy thinking fallback for unsupported GPT-5.6 effort %s",
+      (effort) => {
+        const body = {
+          reasoning: { effort },
+          messages: [{ role: "user", content: "Use legacy thinking" }]
+        };
+
+        const result = openaiToKiroRequest("gpt-5.6-luna", body, true, {});
+
+        expect(result.additionalModelRequestFields).toBeUndefined();
+        expect(systemPromptOf(result)).toContain("<thinking_mode>enabled</thinking_mode>");
+        expect(systemPromptOf(result)).toContain("<max_thinking_length>");
+      }
+    );
+
+    it.each(["none", "off", "disabled"])(
+      "keeps GPT-5.6 reasoning intentionally disabled for effort %s",
+      (effort) => {
+        const body = {
+          reasoning: { effort },
+          messages: [{ role: "user", content: "Do not reason" }]
+        };
+
+        const result = openaiToKiroRequest("gpt-5.6-luna", body, true, {});
+
+        expect(result.additionalModelRequestFields).toBeUndefined();
+        expect(systemPromptOf(result)).not.toContain("<thinking_mode>");
+        expect(systemPromptOf(result)).not.toContain("<max_thinking_length>");
+      }
+    );
+
+    it("keeps the thinking-alias fallback when GPT effort is blank", () => {
+      const body = {
+        reasoning: { effort: "" },
+        messages: [{ role: "user", content: "Use the thinking alias" }]
+      };
+
+      const result = openaiToKiroRequest("gpt-5.6-sol-thinking", body, true, {});
+
+      expect(result.additionalModelRequestFields).toBeUndefined();
+      expect(systemPromptOf(result)).toContain("<thinking_mode>enabled</thinking_mode>");
+      expect(systemPromptOf(result)).toContain("<max_thinking_length>");
+    });
+
     it("does not send additionalModelRequestFields for legacy Kiro model ids", () => {
       const body = {
         reasoning_effort: "high",