diff --git a/changelog.d/fixes/13471-opencode-muse-spark-1-3-responses.md b/changelog.d/fixes/13471-opencode-muse-spark-1-3-responses.md
new file mode 100644
index 0000000000..14aeff3b5d
--- /dev/null
+++ b/changelog.d/fixes/13471-opencode-muse-spark-1-3-responses.md
@@ -0,0 +1 @@
+- **fix(providers):** Muse Spark 1.3 works on OpenCode Zen, OpenCode and OpenCode Go instead of failing with a 500, and gets its real 1M context window ([#13471](https://github.com/diegosouzapw/OmniRoute/pull/13471)) — thanks @maxmad64bis (with thanks to @bacnh85, @shermzy and @atakhadiviom for #12675, #12973 and #13111)
diff --git a/open-sse/config/providers/registry/opencode/go/index.ts b/open-sse/config/providers/registry/opencode/go/index.ts
index 2d364bd5fb..dcc428037f 100644
--- a/open-sse/config/providers/registry/opencode/go/index.ts
+++ b/open-sse/config/providers/registry/opencode/go/index.ts
@@ -226,6 +226,78 @@ export const opencode_goProvider: RegistryEntry = {
       supportsVideo: true,
       targetFormat: "openai-responses",
     },
+    // #12674: Muse Spark 1.3 Contributor — base + effort-tier aliases from the
+    // OpenCode Go registry (`opencode models opencode-go --refresh --verbose`;
+    // exact suffix set: minimal/low/medium/high/xhigh, no max — same as 1.2).
+    // Upstream serves Muse Spark only on the Responses API; without
+    // targetFormat:"openai-responses" these fall through to /chat/completions
+    // and the upstream returns 500 (same class as #12196).
+    {
+      id: "muse-spark-1.3-contributor",
+      name: "Muse Spark 1.3 Contributor",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
+    {
+      id: "muse-spark-1.3-contributor-minimal",
+      name: "Muse Spark 1.3 Contributor (minimal effort)",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
+    {
+      id: "muse-spark-1.3-contributor-low",
+      name: "Muse Spark 1.3 Contributor (low effort)",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
+    {
+      id: "muse-spark-1.3-contributor-medium",
+      name: "Muse Spark 1.3 Contributor (medium effort)",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
+    {
+      id: "muse-spark-1.3-contributor-high",
+      name: "Muse Spark 1.3 Contributor (high effort)",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
+    {
+      id: "muse-spark-1.3-contributor-xhigh",
+      name: "Muse Spark 1.3 Contributor (xhigh effort)",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+      supportsReasoning: true,
+      supportsVision: true,
+      supportsAudio: true,
+      supportsVideo: true,
+      targetFormat: "openai-responses",
+    },
     // #8353: Grok 4.5 + effort tiers from the OpenCode Go registry.
     {
       id: "grok-4.5",
diff --git a/open-sse/config/providers/registry/opencode/index.ts b/open-sse/config/providers/registry/opencode/index.ts
index e402a6e9cf..b667331289 100644
--- a/open-sse/config/providers/registry/opencode/index.ts
+++ b/open-sse/config/providers/registry/opencode/index.ts
@@ -50,6 +50,25 @@ export const opencodeProvider: RegistryEntry = {
       contextLength: 1048576,
       maxOutputTokens: 131072,
     },
+    // Muse Spark 1.3 is served only on the Responses API, same as 1.2 above.
+    // Its window matches the published OpenCode catalog instead of the
+    // 200000 provider default.
+    {
+      id: "muse-spark-1.3",
+      name: "Muse Spark 1.3",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+    },
+    {
+      id: "muse-spark-1.3-contributor-free",
+      name: "Muse Spark 1.3 Contributor Free",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+    },
     { id: "deepseek-v4-flash-free", name: "DeepSeek V4 Flash Free", supportsReasoning: true },
     // #6998: 2026-07-14 refresh — the upstream free tier rotated its lineup;
     // minimax-m3-free, minimax-m2.5-free, ling-2.6-1t-free,
diff --git a/open-sse/config/providers/registry/opencode/zen/index.ts b/open-sse/config/providers/registry/opencode/zen/index.ts
index 75849d65a6..1bb2382aad 100644
--- a/open-sse/config/providers/registry/opencode/zen/index.ts
+++ b/open-sse/config/providers/registry/opencode/zen/index.ts
@@ -85,6 +85,25 @@ export const opencode_zenProvider: RegistryEntry = {
       contextLength: 1048576,
       maxOutputTokens: 131072,
     },
+    // Muse Spark 1.3 is served only on the Responses API, same as 1.2 above.
+    // Its window matches the published OpenCode catalog instead of the
+    // 200000 provider default.
+    {
+      id: "muse-spark-1.3",
+      name: "Muse Spark 1.3",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+    },
+    {
+      id: "muse-spark-1.3-contributor-free",
+      name: "Muse Spark 1.3 Contributor Free",
+      supportsReasoning: true,
+      targetFormat: "openai-responses",
+      contextLength: 1048576,
+      maxOutputTokens: 131072,
+    },
 
     // ── DeepSeek ────────────────────────────────────────────────
     // #10788: same tier vocabulary as opencode-go's DeepSeek rows — the Zen
diff --git a/open-sse/executors/opencode.ts b/open-sse/executors/opencode.ts
index 463c8bec65..ce8ea29b60 100644
--- a/open-sse/executors/opencode.ts
+++ b/open-sse/executors/opencode.ts
@@ -94,6 +94,8 @@ const OPENCODE_FREE_MODELS = new Set([
  *   grok-4.5 low/medium/high; hy3 none/low/high; kimi-k3 max;
  *   qwen3.6-plus / qwen3.7-max / qwen3.7-plus high/max;
  *   muse-spark-1.2-contributor minimal/low/medium/high/xhigh (no max)
+ * - #12674 Muse Spark 1.3 Contributor: minimal/low/medium/high/xhigh (no max),
+ *   verified via `opencode models opencode-go --refresh --verbose`
  */
 const EFFORT_TIERS: Record<string, readonly string[]> = {
   "deepseek-v4-pro": EFFORT_LEVELS,
@@ -107,6 +109,7 @@ const EFFORT_TIERS: Record<string, readonly string[]> = {
   "qwen3.7-max": ["high", "max"],
   "qwen3.7-plus": ["high", "max"],
   "muse-spark-1.2-contributor": ["minimal", "low", "medium", "high", "xhigh"],
+  "muse-spark-1.3-contributor": ["minimal", "low", "medium", "high", "xhigh"],
 };
 
 /**
diff --git a/tests/unit/opencode-go-muse-spark-1-3-12674.test.ts b/tests/unit/opencode-go-muse-spark-1-3-12674.test.ts
new file mode 100644
index 0000000000..bc9f485fb0
--- /dev/null
+++ b/tests/unit/opencode-go-muse-spark-1-3-12674.test.ts
@@ -0,0 +1,75 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import {
+  parseEffortLevel,
+  resolveOpencodeTargetFormat,
+} from "../../open-sse/executors/opencode.ts";
+
+const { REGISTRY } = (await import("../../open-sse/config/providerRegistry.ts")) as {
+  REGISTRY: Record<
+    string,
+    {
+      models?: Array<{
+        id: string;
+        targetFormat?: string;
+        supportsReasoning?: boolean;
+        contextLength?: number;
+        maxOutputTokens?: number;
+      }>;
+    }
+  >;
+};
+
+// #12674: opencode-go/muse-spark-1.3-contributor (upstream release 2026-09-02)
+// 500s via /v1/chat/completions because the upstream serves Muse Spark only on
+// the Responses API and the model had no registry targetFormat entry, so the
+// executor fell back to "openai". Base + effort-tier alias set verified via
+// `opencode models opencode-go --refresh --verbose` (minimal/low/medium/high/
+// xhigh, no max — same as 1.2).
+const BASE = "muse-spark-1.3-contributor";
+const ALIASES = ["minimal", "low", "medium", "high", "xhigh"].map((effort) => ({
+  alias: `${BASE}-${effort}`,
+  effort,
+}));
+
+function goModels() {
+  const entry = REGISTRY["opencode-go"];
+  assert.ok(entry, "opencode-go registry entry must exist");
+  return entry.models ?? [];
+}
+
+test("#12674 registry: 1.3 base + effort aliases target the Responses API", () => {
+  const models = goModels();
+  for (const id of [BASE, ...ALIASES.map((a) => a.alias)]) {
+    const model = models.find((m) => m.id === id);
+    assert.ok(model, `${id} must be registered on opencode-go`);
+    assert.equal(model?.targetFormat, "openai-responses", `${id} must target Responses`);
+    assert.equal(model?.supportsReasoning, true, `${id} must support reasoning`);
+    assert.equal(model?.contextLength, 1048576, `${id} context must be 1M`);
+    assert.equal(model?.maxOutputTokens, 131072, `${id} max output must be 128K`);
+  }
+});
+
+test("#12674 executor: 1.3 resolves to openai-responses (URL selection)", () => {
+  assert.equal(resolveOpencodeTargetFormat("opencode-go", BASE), "openai-responses");
+  assert.equal(resolveOpencodeTargetFormat("opencode-go", `${BASE}-high`), "openai-responses");
+});
+
+for (const { alias, effort } of ALIASES) {
+  test(`#12674 parseEffortLevel: ${alias} → ${effort}`, () => {
+    assert.deepEqual(parseEffortLevel(alias), { baseModel: BASE, effort });
+  });
+}
+
+test("#12674 parseEffortLevel: 1.3 has no max tier", () => {
+  assert.equal(parseEffortLevel(`${BASE}-max`), null);
+});
+
+test("#12674 catalog: 1.3 aliases stay Go-only (not on opencode-zen)", () => {
+  const zenIds = new Set((REGISTRY["opencode-zen"]?.models ?? []).map((m) => m.id));
+  assert.equal(zenIds.has(BASE), false, "opencode-zen must not expose 1.3 base");
+  for (const { alias } of ALIASES) {
+    assert.equal(zenIds.has(alias), false, `opencode-zen must not expose ${alias}`);
+  }
+});
diff --git a/tests/unit/opencode-muse-spark-1-3-responses-12674.test.ts b/tests/unit/opencode-muse-spark-1-3-responses-12674.test.ts
new file mode 100644
index 0000000000..fcb70fdadf
--- /dev/null
+++ b/tests/unit/opencode-muse-spark-1-3-responses-12674.test.ts
@@ -0,0 +1,90 @@
+import assert from "node:assert/strict";
+import test from "node:test";
+
+import { opencode_zenProvider } from "../../open-sse/config/providers/registry/opencode/zen/index.ts";
+import { opencodeProvider } from "../../open-sse/config/providers/registry/opencode/index.ts";
+import { getTokenLimit } from "../../open-sse/services/contextManager.ts";
+import {
+  OpencodeExecutor,
+  resolveOpencodeTargetFormat,
+} from "../../open-sse/executors/opencode.ts";
+
+// Issue 12674: Muse Spark 1.3 is served on the Responses API with a 1M
+// context window. These entries only set the wire format; without the
+// explicit window the limit fell back to the 200000 provider default.
+const PROVIDERS = [
+  { id: "opencode-zen", entry: opencode_zenProvider },
+  { id: "opencode", entry: opencodeProvider },
+] as const;
+const MODEL_IDS = ["muse-spark-1.3", "muse-spark-1.3-contributor-free"] as const;
+
+test("muse-spark-1.3 models target the Responses API with reasoning enabled", () => {
+  for (const { id, entry } of PROVIDERS) {
+    for (const modelId of MODEL_IDS) {
+      const model = entry.models.find((m) => m.id === modelId);
+      assert.ok(model, `${modelId} should be registered on ${id}`);
+      assert.equal(
+        model?.targetFormat,
+        "openai-responses",
+        `${modelId} on ${id} must target the Responses API, not the default chat/completions pass-through`
+      );
+      assert.equal(model?.supportsReasoning, true);
+    }
+  }
+});
+
+test("muse-spark-1.3 models declare the real 1M context window and 128K output limit", () => {
+  for (const { id, entry } of PROVIDERS) {
+    for (const modelId of MODEL_IDS) {
+      const model = entry.models.find((m) => m.id === modelId);
+      assert.ok(model, `${modelId} should be registered on ${id}`);
+      assert.equal(model?.contextLength, 1048576, `${modelId} on ${id} must declare a 1M window`);
+      assert.equal(
+        model?.maxOutputTokens,
+        131072,
+        `${modelId} on ${id} must declare a 128K output limit`
+      );
+    }
+  }
+});
+
+test("context limit resolves muse-spark-1.3 models to the 1M window, not the provider default", () => {
+  for (const { id } of PROVIDERS) {
+    for (const modelId of MODEL_IDS) {
+      assert.equal(
+        getTokenLimit(id, modelId),
+        1048576,
+        `getTokenLimit(${id}, ${modelId}) must return the 1M window`
+      );
+    }
+  }
+});
+
+test("executor resolves muse-spark-1.3 models to the Responses API", () => {
+  for (const { id } of PROVIDERS) {
+    for (const modelId of MODEL_IDS) {
+      assert.equal(
+        resolveOpencodeTargetFormat(id, modelId),
+        "openai-responses",
+        `resolveOpencodeTargetFormat(${id}, ${modelId}) must return openai-responses`
+      );
+    }
+  }
+});
+
+test("Responses-format executor hits /responses with x-api-key auth and no Bearer header", () => {
+  for (const { id } of PROVIDERS) {
+    for (const modelId of MODEL_IDS) {
+      const executor = new OpencodeExecutor(id);
+      executor._requestFormat = resolveOpencodeTargetFormat(id, modelId);
+      const url = executor.buildUrl(modelId, true);
+      assert.ok(
+        url.endsWith("/responses"),
+        `${id}/${modelId} must build a /responses URL, got ${url}`
+      );
+      const headers = executor.buildHeaders({ apiKey: "sk-test" }, true, null, modelId);
+      assert.equal(headers["x-api-key"], "sk-test");
+      assert.equal(headers["Authorization"], undefined);
+    }
+  }
+});