diff --git a/changelog.d/fixes/6721-codex-spark-image-drop.md b/changelog.d/fixes/6721-codex-spark-image-drop.md
new file mode 100644
index 0000000000..06cc548828
--- /dev/null
+++ b/changelog.d/fixes/6721-codex-spark-image-drop.md
@@ -0,0 +1 @@
+- **fix(providers):** Codex Desktop requests to `gpt-5.3-codex-spark` failed with `[400]: Tool 'image_generation' is not supported with gpt-5.3-codex-spark`, even on paid-plan accounts ([#6651](https://github.com/diegosouzapw/OmniRoute/issues/6651)) — `CodexExecutor.transformRequest` (`open-sse/executors/codex.ts`) only dropped the Codex Desktop-injected `image_generation` hosted tool when `isCodexFreePlan()` matched the account's plan, with no awareness that Spark-scope Codex models reject `image_generation` upstream regardless of plan. `dropImageGeneration` now also drops it when `getCodexModelScope(model) === "spark"` (the existing Spark classifier from `open-sse/config/codexQuotaScopes.ts`), independent of account plan. Regression guard: `tests/unit/codex-spark-image-generation.test.ts`.
diff --git a/open-sse/executors/codex.ts b/open-sse/executors/codex.ts
index ceaa8bda69..2fd66b2343 100644
--- a/open-sse/executors/codex.ts
+++ b/open-sse/executors/codex.ts
@@ -1132,7 +1132,11 @@ export class CodexExecutor extends BaseExecutor {
     // Cursor may include custom tools (e.g. ApplyPatch) that work locally but are
     // invalid upstream, and translation bugs can leave orphaned/empty tool_choice names.
     normalizeCodexTools(body, {
-      dropImageGeneration: isCodexFreePlan(credentials?.providerSpecificData),
+      // gpt-5.3-codex-spark (and other Spark-scope models) reject image_generation
+      // upstream even on paid-plan accounts, so drop it independent of plan (#6651).
+      dropImageGeneration:
+        isCodexFreePlan(credentials?.providerSpecificData) ||
+        getCodexModelScope(model) === "spark",
       preserveCustomTools: nativeCodexPassthrough,
     });
 
diff --git a/tests/unit/codex-spark-image-generation.test.ts b/tests/unit/codex-spark-image-generation.test.ts
new file mode 100644
index 0000000000..7d2671b866
--- /dev/null
+++ b/tests/unit/codex-spark-image-generation.test.ts
@@ -0,0 +1,61 @@
+/**
+ * #6651 — Codex Desktop injects the `image_generation` hosted tool into every
+ * Responses API request. OmniRoute only dropped it for free-plan Codex
+ * accounts (isCodexFreePlan). It did NOT drop it for gpt-5.3-codex-spark (and
+ * other Spark-scope models), which reject `image_generation` upstream even on
+ * paid-plan accounts, producing:
+ * [400]: Tool 'image_generation' is not supported with gpt-5.3-codex-spark.
+ *
+ * Fix: CodexExecutor.transformRequest now also drops image_generation when
+ * the target model resolves to the Spark quota scope
+ * (getCodexModelScope(model) === "spark"), independent of plan.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+const { CodexExecutor } = await import("../../open-sse/executors/codex.ts");
+
+function buildBody() {
+  return {
+    _nativeCodexPassthrough: true,
+    tools: [
+      { type: "image_generation", output_format: "png" },
+      { type: "function", name: "foo", parameters: { type: "object" } },
+    ],
+  };
+}
+
+test("#6651: CodexExecutor.transformRequest drops image_generation for gpt-5.3-codex-spark even on a paid-plan account", () => {
+  const executor = new CodexExecutor();
+
+  // Paid-plan account (not free) — isCodexFreePlan() alone returns false, so
+  // the fix must rely on the model-scope check to still drop the tool.
+  const result = executor.transformRequest("gpt-5.3-codex-spark", buildBody(), false, {
+    providerSpecificData: { workspacePlanType: "team" },
+  }) as { tools: Array<{ type?: string }> };
+
+  assert.equal(
+    result.tools.some((t) => t.type === "image_generation"),
+    false,
+    "image_generation must be dropped for gpt-5.3-codex-spark regardless of account plan (#6651)"
+  );
+  assert.equal(
+    result.tools.some((t) => t.type === "function"),
+    true,
+    "the function tool must survive"
+  );
+});
+
+test("#6651: CodexExecutor.transformRequest still preserves image_generation for non-Spark models on paid plans", () => {
+  const executor = new CodexExecutor();
+
+  const result = executor.transformRequest("gpt-5", buildBody(), false, {
+    providerSpecificData: { workspacePlanType: "team" },
+  }) as { tools: Array<{ type?: string }> };
+
+  assert.equal(
+    result.tools.some((t) => t.type === "image_generation"),
+    true,
+    "image_generation must still be preserved for non-Spark models on paid plans"
+  );
+});