From 0e2d04526c6b08b8a7bf2e2e16e6cdbbf1f5a21d Mon Sep 17 00:00:00 2001
From: terence71-glitch <mcdowellterence71@gmail.com>
Date: Mon, 18 May 2026 00:38:25 +0200
Subject: [PATCH] fix: honor Codex reasoning suffix aliases (#2335)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Integrated into release/v3.8.0 — fixes issue #2331: Codex model suffix aliases now correctly override client-injected reasoning defaults
---
 open-sse/executors/codex.ts | 12 +++++-------
 tests/unit/plan3-p0.test.ts | 19 +++++++++++++++++++
 2 files changed, 24 insertions(+), 7 deletions(-)

diff --git a/open-sse/executors/codex.ts b/open-sse/executors/codex.ts
index 5351677789..eb207f2683 100644
--- a/open-sse/executors/codex.ts
+++ b/open-sse/executors/codex.ts
@@ -1388,15 +1388,13 @@ export class CodexExecutor extends BaseExecutor {
     const fallbackReasoningEffort = allowConnectionReasoningDefaults
       ? requestDefaults.reasoningEffort || "medium"
       : undefined;
+    // Issue #2331: model suffix aliases (for example gpt-5.5-xhigh) represent an
+    // explicit model selection, so they must override client-injected defaults such
+    // as OpenCode's automatic reasoning.effort=medium for GPT-5-family requests.
     const rawEffort =
-      explicitReasoning || requestReasoningEffort || modelEffort || fallbackReasoningEffort;
+      modelEffort || explicitReasoning || requestReasoningEffort || fallbackReasoningEffort;
 
-    if (explicitReasoning) {
-      body.reasoning = {
-        ...(reasoningRecord || {}),
-        effort: clampEffort(cleanModel, explicitReasoning),
-      };
-    } else if (rawEffort) {
+    if (rawEffort) {
       body.reasoning = {
         ...(reasoningRecord || {}),
         effort: clampEffort(cleanModel, rawEffort),
diff --git a/tests/unit/plan3-p0.test.ts b/tests/unit/plan3-p0.test.ts
index 7c4a786dcf..d622861fe9 100644
--- a/tests/unit/plan3-p0.test.ts
+++ b/tests/unit/plan3-p0.test.ts
@@ -317,6 +317,25 @@ test("CodexExecutor preserves native responses payloads for Codex passthrough",
   assert.ok(!("_nativeCodexPassthrough" in transformed));
 });
 
+test("CodexExecutor gives model reasoning suffix precedence over client defaults", () => {
+  const executor = new CodexExecutor();
+  const transformed = executor.transformRequest(
+    "gpt-5.5-xhigh",
+    {
+      model: "gpt-5.5-xhigh",
+      input: [],
+      reasoning: { effort: "medium", summary: "auto" },
+      reasoning_effort: "low",
+    },
+    true,
+    {}
+  );
+
+  assert.equal(transformed.model, "gpt-5.5");
+  assert.deepEqual(transformed.reasoning, { effort: "xhigh", summary: "auto" });
+  assert.equal(transformed.reasoning_effort, undefined);
+});
+
 test("CodexExecutor strips streaming fields for compact passthrough", () => {
   const executor = new CodexExecutor();
   const transformed = executor.transformRequest(