MaxGhenis · Oct 13, 2026 · Oct 12, 2026 · Oct 12, 2026 · Oct 12, 2026 · Oct 12, 2026
diff --git a/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_adjudications.json b/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_adjudications.json
diff --git a/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_audit_row_annotations.csv b/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_audit_row_annotations.csv
diff --git a/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_case_notes.csv b/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_case_notes.csv
diff --git a/...ions/us_full_run_20260612_policyengine_4_16_1_populace/us_case_reference_explanations.csv b/...ions/us_full_run_20260612_policyengine_4_16_1_populace/us_case_reference_explanations.csv
diff --git a/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_wording-amendments.json b/annotations/us_full_run_20260612_policyengine_4_16_1_populace/us_wording-amendments.json
diff --git a/app/public/paper/policybench.pdf b/app/public/paper/policybench.pdf
diff --git a/app/public/paper/web/figures/positive_zero_scatter.png b/app/public/paper/web/figures/positive_zero_scatter.png
diff --git a/app/public/paper/web/index.html b/app/public/paper/web/index.html
diff --git a/app/src/components/ProviderMark.tsx b/app/src/components/ProviderMark.tsx
@@ -3,6 +3,7 @@ import {
   Anthropic,
   DeepSeek,
   Google,
+  Meta,
   Minimax,
   Moonshot,
   OpenAI,
@@ -78,6 +79,7 @@ export default function ProviderMark({
       {provider === "anthropic" && <Anthropic size={size} color="currentColor" aria-hidden="true" />}
       {provider === "deepseek" && <DeepSeek.Color size={size} aria-hidden="true" />}
       {provider === "google" && <Google.Color size={size} aria-hidden="true" />}
+      {provider === "meta" && <Meta.Color size={size} aria-hidden="true" />}
       {provider === "minimax" && <Minimax.Color size={size} aria-hidden="true" />}
       {provider === "moonshot" && <Moonshot size={size} color="currentColor" aria-hidden="true" />}
       {provider === "openai" && <OpenAI size={size} color="currentColor" aria-hidden="true" />}

diff --git a/app/src/data.artifact.json b/app/src/data.artifact.json
@@ -1,9 +1,9 @@
 {
   "asset": "dashboard-data.json",
-  "bytes": 128756688,
+  "bytes": 130988798,
   "repo": "PolicyEngine/policybench",
-  "sha256": "f834478e65191f8426ef479506f9f6fe32c2e7107f30fdbf4e7ba0e6acba6410",
-  "tag": "dashboard-data-20261010",
-  "url": "https://github.com/PolicyEngine/policybench/releases/download/dashboard-data-20261010/dashboard-data.json",
+  "sha256": "aef5dc73a76433c97994430f8eb6dfad6bdc6518b18ee3ea8f79cee704757477",
+  "tag": "dashboard-data-20261013",
+  "url": "https://github.com/PolicyEngine/policybench/releases/download/dashboard-data-20261013/dashboard-data.json",
   "version": 1
 }
diff --git a/app/src/data.versions.json b/app/src/data.versions.json
@@ -6,10 +6,10 @@
       "artifact": {
         "pointer": "live"
       },
-      "description": "Scored reference outputs from policyengine-us 2.38.6 (the 58 excluded outputs keep the values they were decided on: 38 from policyengine-us 1.755.4, 18 from 2.15.17, 2 from 2.38.6, and the 17 that move on 2.38.6 were re-reviewed and stay excluded); same household facts and requested outputs for every model; answer transport and request shape per model as recorded in the serving-configuration table - 47 models",
+      "description": "Scored reference outputs from policyengine-us 2.41.0 (the 49 excluded outputs keep the values they were decided on: 29 from policyengine-us 1.755.4, 18 from 2.15.17, 2 from 2.38.6, and the 12 that move on 2.41.0 were re-reviewed and stay excluded); same household facts and requested outputs for every model; answer transport and request shape per model as recorded in the serving-configuration table - 48 models",
       "id": "1.1",
       "label": "1.1",
-      "snapshotLabel": "Snapshot 2026-10-10"
+      "snapshotLabel": "Snapshot 2026-10-13"
     },
     {
       "artifact": {

diff --git a/app/src/lib/referenceEngine.ts b/app/src/lib/referenceEngine.ts
@@ -15,8 +15,8 @@ import type { ReferenceExclusion } from "../types";
  * the frozen sidecar.
  */
 export const ENGINE_UPGRADE_RECHECK = {
-  engineVersion: "2.38.6",
-  rechecked: 17,
+  engineVersion: "2.41.0",
+  rechecked: 12,
 };
 
 function compareVersions(a: string, b: string): number {

diff --git a/app/src/lib/servingSensitivity.ts b/app/src/lib/servingSensitivity.ts
@@ -7,7 +7,9 @@
  * reasoning in both runs, so its comparison is one of transport. Claude Opus
  * 5.5 and Claude Sonnet 5.5 reject forced calls the same way and have no auto
  * re-run. Claude Haiku 5.5 accepts forced calls, so its board row is a
- * forced-tool row like Claude Opus 5's. Scores are
+ * forced-tool row like Claude Opus 5's. Meta's Muse Wonder, not a Claude row,
+ * rejects forced calls too and reasons on JSON, so it has no re-run either.
+ * Scores are
  * the pinned three-decimal measurements from sensitivity/data/*.json; round
  * only for display. Each entry's "would rank" is derived from the live board
  * rows at render time, never typed by hand.
@@ -36,30 +38,30 @@ const FORCED_TOOL =
 
 export const SERVING_SENSITIVITY: Record<string, ServingSensitivity> = {
   "claude-fable-5": {
-    autoExact: 91.746,
+    autoExact: 91.451,
     boardTreatment: `${FORCED_TOOL}, one output per request`,
     autoTreatment:
       "re-run with tool_choice: auto and the whole household in one request",
     thinkingSuppressedOnBoard: true,
     noteHref: SENSITIVITY_DOC_HREF,
   },
   "claude-opus-5": {
-    autoExact: 90.456,
+    autoExact: 90.109,
     boardTreatment: FORCED_TOOL,
     autoTreatment: "re-run with tool_choice: auto",
     thinkingSuppressedOnBoard: true,
     noteHref: SENSITIVITY_DOC_HREF,
   },
   "claude-sonnet-5": {
-    autoExact: 84.953,
+    autoExact: 84.574,
     boardTreatment: `${FORCED_TOOL}, one output per request`,
     autoTreatment:
       "re-run with tool_choice: auto and the whole household in one request",
     thinkingSuppressedOnBoard: true,
     noteHref: SENSITIVITY_DOC_HREF,
   },
   "claude-fable-5.1": {
-    autoExact: 92.044,
+    autoExact: 91.748,
     boardTreatment:
       "rejects forced tool calls, so its row answers as a JSON object and reasons at the provider default",
     autoTreatment:
@@ -68,7 +70,7 @@ export const SERVING_SENSITIVITY: Record<string, ServingSensitivity> = {
     noteHref: "/notes/2026-09-01-claude-fable-5-1-added",
   },
   "claude-haiku-5.5": {
-    autoExact: 89.239,
+    autoExact: 88.82,
     boardTreatment: FORCED_TOOL,
     autoTreatment: "re-run with tool_choice: auto",
     thinkingSuppressedOnBoard: true,

diff --git a/app/src/model-serving-config.json b/app/src/model-serving-config.json
@@ -13,7 +13,7 @@
   },
   "evidence_summary": {
     "registry": 29,
-    "run_state": 18
+    "run_state": 19
   },
   "models": {
     "claude-fable-5": {
@@ -1493,6 +1493,63 @@
       "shared_completion_budget_tokens": 16384,
       "tool_choice": "forced"
     },
+    "muse-wonder": {
+      "answer_contract": "json",
+      "evidence": {
+        "compared_columns": [
+          "scenario_id",
+          "variable",
+          "prediction",
+          "explanation",
+          "raw_response",
+          "provider_resolved_model",
+          "prompt_tokens",
+          "completion_tokens",
+          "error"
+        ],
+        "fields": [
+          "answer_contract",
+          "chunk_size",
+          "completion_budget_ceiling",
+          "fingerprint_version",
+          "initial_completion_budget_tokens",
+          "max_repair_rounds",
+          "model_id",
+          "prompt_contract_version",
+          "request_timeout_seconds",
+          "thinking",
+          "tool_choice_mode"
+        ],
+        "frozen_rows_sha256": "ed4cbd03e82d4874de4243dc454922f4851ff13b53d59213cc34e9ad0870485a",
+        "kind": "run_state",
+        "rows_match": true,
+        "run": "musewonder-runs/musewonder",
+        "source_predictions_sha256": "2e0eddd68fdd39201f564ed07d485329e13a80450881a3bad63d24f726380d6a",
+        "source_rows_sha256": "ed4cbd03e82d4874de4243dc454922f4851ff13b53d59213cc34e9ad0870485a",
+        "treatment_fingerprint": {
+          "answer_contract": "json",
+          "chunk_size": null,
+          "completion_budget_ceiling": 128000,
+          "fingerprint_version": 3,
+          "initial_completion_budget_tokens": 16384,
+          "max_repair_rounds": 2,
+          "model_id": "openrouter/meta/muse-wonder",
+          "prompt_contract_version": "2026-08-09-v2-scoring-contract",
+          "request_timeout_seconds": 300,
+          "thinking": {
+            "mode": "provider_default"
+          },
+          "tool_choice_mode": null
+        }
+      },
+      "provider_id": "openrouter/meta/muse-wonder",
+      "reasoning_setup": "provider default; 16,384-token shared budget",
+      "registry_derived": [],
+      "request_shape": "whole scenario",
+      "request_timeout_seconds": 300,
+      "shared_completion_budget_tokens": 16384,
+      "tool_choice": null
+    },
     "ox-alpha": {
       "answer_contract": "tool",
       "evidence": {
@@ -1585,7 +1642,7 @@
       "tool_choice": null
     }
   },
-  "registry_commit": "0562d77a9d263c14bb9c9fa93ba3c765fcff95a9",
+  "registry_commit": "3ef45909db7367abdffd08379babde5bcf5c549e",
   "sources": [
     "policybench.config.MODELS",
     "policybench.model_cards",

diff --git a/app/src/modelMeta.ts b/app/src/modelMeta.ts
@@ -52,6 +52,7 @@ export const MODEL_ORDER = [
   "inkling",
   "qwen3.8-max",
   "qwen-3.7-max",
+  "muse-wonder",
 ] as const;
 
 export const MODEL_LABELS: Record<string, string> = {
@@ -103,13 +104,15 @@ export const MODEL_LABELS: Record<string, string> = {
   "qwen-3.7-max": "Qwen 3.7 Max",
   "qwen3.8-max": "Qwen 3.8 Max",
   inkling: "Inkling",
+  "muse-wonder": "Muse Wonder",
 };
 
 export type ProviderKey =
   | "alibaba"
   | "anthropic"
   | "deepseek"
   | "google"
+  | "meta"
   | "minimax"
   | "moonshot"
   | "openai"
@@ -177,6 +180,7 @@ export const PROVIDER_LABELS: Record<ProviderKey, string> = {
   anthropic: "Anthropic",
   deepseek: "DeepSeek",
   google: "Google",
+  meta: "Meta",
   minimax: "MiniMax",
   moonshot: "Moonshot AI",
   openai: "OpenAI",
@@ -196,6 +200,7 @@ export function getProviderForModel(model: string): ProviderKey | null {
   if (model.startsWith("grok-")) return "xai";
   if (model.startsWith("kimi-")) return "moonshot";
   if (model.startsWith("minimax-")) return "minimax";
+  if (model.startsWith("muse-")) return "meta";
   if (model.startsWith("qwen")) return "alibaba";
   if (model.startsWith("inkling")) return "thinkingmachines";
   return null;
@@ -227,6 +232,7 @@ const FRONTIER_MODEL_GROUPS = [
   ["minimax-m3"],
   ["inkling"],
   ["qwen3.8-max", "qwen-3.7-max"],
+  ["muse-wonder"],
 ] as const;
 
 export const FRONTIER_MODELS: readonly string[] = FRONTIER_MODEL_GROUPS.map(

diff --git a/app/src/notes/2026-10-13-muse-wonder-joins-the-board.json b/app/src/notes/2026-10-13-muse-wonder-joins-the-board.json
@@ -0,0 +1,62 @@
+{
+  "slug": "2026-10-13-muse-wonder-joins-the-board",
+  "date": "2026-10-13",
+  "title": "Meta's Muse Wonder joins the board as the references move to policyengine-us 2.41.0",
+  "paragraphs": [
+    "Meta's Muse Wonder joined the board on {releaseDate}, the day after Meta released it, bringing the board to {nModels} models. It is the first Meta model on the board. It gets {museExact}% of answers within $1, weighted by household impact, #{museRank} of {nModels}. On the bounded score, which gives partial credit for amounts close to the reference, it ranks #{museBoundedRank}, behind only GPT-6 Sol and Claude Opus 5.5. GPT-6 Sol still leads on answers within $1 at {solExact}%, ahead of Claude Opus 5.5 ({opusExact}%) and GPT-5.6 Sol ({sol56Exact}%).",
+    "Meta released Muse Wonder Full the same day, but only to Muse Code subscribers, with no API. PolicyBench runs every model through its provider's API under one protocol, so Muse Wonder Full is not on the board. PolicyBench reaches Muse Wonder through OpenRouter, whose one endpoint for it is Meta's own API. That API accepts only an automatic tool choice and rejects a forced tool call, so the row answers as a JSON object, as Claude Opus 5.5's and Claude Sonnet 5.5's do, and reasons at Meta's default setting.",
+    "The release also moves PolicyBench's references from policyengine-us {previousEngine} to {engineVersion}, the first release with every fix the previous release was waiting on. Of the outputs the previous release excluded, {restored:words} return to scoring at the new version's values, each within $1 of the corrected value PolicyBench's audits computed for it. Between them they take the upstream fixes for {restoredFixes}. The new version also fixes the defect behind a SNAP output, the heat-and-eat utility allowance, but that output stays excluded because the stated facts do not decide it. No reference PolicyBench already scored changes, and nothing is newly excluded. Every model is now scored on {scoredOutputs} of its {totalOutputs} requested outputs; PolicyBench excludes the other {excluded}.",
+    "Against the previous release, the exact rate of every one of the {incumbents} earlier models falls, by {fallMin} to {fallMax} points. The cause is the {restored:words} restored outputs, which few models get right: half of them are answered within $1 by at most {restoredHalfAtMost:words} of the {nModels} models, and {restoredUnmatched:words} by none. Muse Wonder answers {museRestored:words} of them within $1. Without them, every earlier model's rate would be the previous release's."
+  ],
+  "facts": {
+    "releaseDate": "2026-10-13",
+    "nModels": 48,
+    "museExact": 90.1,
+    "museRank": 10,
+    "museBoundedRank": 3,
+    "solExact": 95.2,
+    "opusExact": 94.1,
+    "sol56Exact": 93.1,
+    "previousEngine": "2.38.6",
+    "engineVersion": "2.41.0",
+    "restored": 9,
+    "restoredFixes": "Idaho's subtraction for health insurance premiums (policyengine-us #10023; two outputs), Wisconsin's retirement income exclusion before refundable credits and Wisconsin's capital gain subtraction for distributions (policyengine-us #10026; two outputs), the IRA deduction's active-participant phase-out, California's itemized deduction conformity and California's educator expense addback (policyengine-us #10032, #10034, #10052) and elective deferrals in the earned income behind the EITC and the refundable child tax credit (policyengine-us #10030; four outputs)",
+    "scoredOutputs": 1935,
+    "totalOutputs": 1984,
+    "excluded": 49,
+    "incumbents": 47,
+    "fallMin": 0.18,
+    "fallMax": 0.42,
+    "restoredHalfAtMost": 1,
+    "restoredUnmatched": 4,
+    "museRestored": 1
+  },
+  "data": [
+    {
+      "label": "Dashboard data release",
+      "href": "https://github.com/PolicyEngine/policybench/releases/tag/dashboard-data-20261013"
+    },
+    {
+      "label": "Engine upgrade record",
+      "href": "https://github.com/PolicyEngine/policybench/blob/main/reference_audit/2026-10-12-engine-upgrade"
+    },
+    {
+      "label": "Reference sidecar with every change",
+      "href": "https://github.com/PolicyEngine/policybench/blob/main/paper/snapshot/20260501/runs/us_full_run_20260612_policyengine_4_16_1_populace/reference_outputs.csv.meta.json"
+    },
+    {
+      "label": "Exclusion record",
+      "href": "https://github.com/PolicyEngine/policybench/blob/main/paper/snapshot/20260501/runs/us_full_run_20260612_policyengine_4_16_1_populace/reference_exclusions.json"
+    },
+    {
+      "label": "Muse Wonder model page",
+      "href": "/model/muse-wonder"
+    },
+    {
+      "label": "Previous release note (October 10)",
+      "href": "/notes/2026-10-10-claude-haiku-5-5-joins-the-board"
+    }
+  ],
+  "boardSnapshot": "2026-10-13",
+  "release": "dashboard-data-20261013"
+}
diff --git a/app/src/notes/index.ts b/app/src/notes/index.ts
@@ -1,3 +1,4 @@
+import museWonderJoinsTheBoard from "./2026-10-13-muse-wonder-joins-the-board.json";
 import claudeHaiku55JoinsTheBoard from "./2026-10-10-claude-haiku-5-5-joins-the-board.json";
 import claudeFable51Added from "./2026-09-01-claude-fable-5-1-added.json";
 import claudeSonnet55DebutsFifth from "./2026-09-29-claude-sonnet-5-5-debuts-fifth.json";
@@ -30,6 +31,7 @@ export type PolicyBenchNote = {
 };
 
 export const notes = [
+  museWonderJoinsTheBoard,
   claudeHaiku55JoinsTheBoard,
   stopsScoringEightTaxOutputs,
   fiveSnapHouseholdsBbce,

diff --git a/app/src/paperSnapshot.json b/app/src/paperSnapshot.json
@@ -1,6 +1,6 @@
 {
-  "pdfVersion": "1607c71abaf3",
-  "responseWindow": "June 12 and October 8, 2026",
-  "snapshotDate": "2026-10-10",
-  "webVersion": "31cd4141373a"
+  "pdfVersion": "88e3c81e3e13",
+  "responseWindow": "June 12 and October 13, 2026",
+  "snapshotDate": "2026-10-13",
+  "webVersion": "af35d8e23eaf"
 }
diff --git a/app/tests/auditUniverse.test.ts b/app/tests/auditUniverse.test.ts
@@ -18,17 +18,17 @@ describe("audit universe", () => {
     expect(AUDIT_SELECTION_RULE).toBe(
       "rows whose legacy threshold score is below 1",
     );
-    // Release 20261010: the counts tests/test_disclosures.py recomputes
-    // from the frozen annotations and payload (47 models; the ten outputs
-    // ruled on 2026-10-06 and the two Indiana county outputs leave scoring,
-    // the outputs the engine upgrade regenerates return to it).
+    // Release 20261013: the counts tests/test_disclosures.py recomputes
+    // from the frozen annotations and payload (48 models; Muse Wonder's rows
+    // join, and the nine outputs the 2.41.0 upgrade regenerates return to
+    // scoring).
     expect(summarizeAuditUniverse(us)).toEqual({
-      annotatedRowCount: 8_001,
-      legacyThresholdRowCount: 8_001,
-      exactMissCount: 7_997,
-      annotatedExactMissCount: 7_997,
+      annotatedRowCount: 8_503,
+      legacyThresholdRowCount: 8_503,
+      exactMissCount: 8_499,
+      annotatedExactMissCount: 8_499,
       annotatedExactHitCount: 4,
-      unannotatedBelowFullBoundedScoreCount: 2_138,
+      unannotatedBelowFullBoundedScoreCount: 2_218,
     });
   });
 });
diff --git a/app/tests/expandPage.test.ts b/app/tests/expandPage.test.ts
@@ -71,18 +71,19 @@ test("expand page derives its exact-score headline from the live summary", async
   ).text();
 
   expect(leader.model).toBe("gpt-6-sol");
-  expect(leader.exact).toBeCloseTo(95.583, 3);
+  expect(leader.exact).toBeCloseTo(95.225, 3);
   expect(pageSource).toContain("const leader = headlineExactLeader(dashboard);");
   expect(pageSource).toContain(
     "const leaderLabel = MODEL_LABELS[leader.model] ?? leader.model;",
   );
   expect(pageSource).toContain("The best model, {leaderLabel}, computes");
   expect(pageSource).toContain("{leader.exact.toFixed(1)}% of");
+  expect(pageSource).not.toContain("95.2%");
   expect(pageSource).not.toContain("95.6%");
   expect(pageSource).not.toContain("96.5%");
   expect(pageSource).not.toContain("95.0%");
   expect(expectedCopy).toBe(
-    "GPT-6 Sol computes 95.6% of requested outputs exactly",
+    "GPT-6 Sol computes 95.2% of requested outputs exactly",
   );
 });
 

diff --git a/app/tests/heatmapMetric.test.ts b/app/tests/heatmapMetric.test.ts
@@ -39,8 +39,8 @@ describe("program heatmap metric", () => {
     expect(Math.round(heatmapValue(federal, "exact"))).toBe(80);
     expect(Math.round(heatmapValue(federal, "score"))).toBe(96);
     const stateTax = fable51("state_income_tax_before_refundable_credits");
-    expect(Math.round(heatmapValue(stateTax, "exact"))).toBe(74);
-    expect(Math.round(heatmapValue(stateTax, "score"))).toBe(94);
+    expect(Math.round(heatmapValue(stateTax, "exact"))).toBe(69);
+    expect(Math.round(heatmapValue(stateTax, "score"))).toBe(92);
   });
 
   test("the rendered heatmap shows exact-match cells and says so", () => {