MaxGhenis · Oct 14, 2026 · Sep 30, 2026 · Sep 30, 2026 · Sep 30, 2026 · Sep 30, 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/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/data.artifact.json b/app/src/data.artifact.json
@@ -1,9 +1,9 @@
 {
   "asset": "dashboard-data.json",
-  "bytes": 130982477,
+  "bytes": 135840397,
   "repo": "PolicyEngine/policybench",
-  "sha256": "0a4b754a4ce3b249e0608efac62e52646e4c6f33c6a28c1a496c4c9fda49b507",
-  "tag": "dashboard-data-20261013b",
-  "url": "https://github.com/PolicyEngine/policybench/releases/download/dashboard-data-20261013b/dashboard-data.json",
+  "sha256": "22db9a9c052545371687f6afa3551896336abbd15c76f767ebb777a77f6d2268",
+  "tag": "dashboard-data-20261014",
+  "url": "https://github.com/PolicyEngine/policybench/releases/download/dashboard-data-20261014/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 2.41.0 (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",
+      "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 - 50 models",
       "id": "1.1",
       "label": "1.1",
-      "snapshotLabel": "Snapshot 2026-10-13"
+      "snapshotLabel": "Snapshot 2026-10-14"
     },
     {
       "artifact": {

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": 19
+    "run_state": 21
   },
   "models": {
     "claude-fable-5": {
@@ -820,6 +820,63 @@
       "shared_completion_budget_tokens": 16384,
       "tool_choice": "forced"
     },
+    "gemini-4-argon": {
+      "answer_contract": "tool",
+      "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": "938c3f632d467dcf2cb97391265df103fba104633ac6a10a0ee6c9fe750c8335",
+        "kind": "run_state",
+        "rows_match": true,
+        "run": "argon-runs/argon",
+        "source_predictions_sha256": "76724f96baef92d1958cfdbdcbbaec82b4a06d6b72c0e891534a9fa03e156462",
+        "source_rows_sha256": "938c3f632d467dcf2cb97391265df103fba104633ac6a10a0ee6c9fe750c8335",
+        "treatment_fingerprint": {
+          "answer_contract": "tool",
+          "chunk_size": null,
+          "completion_budget_ceiling": 32768,
+          "fingerprint_version": 3,
+          "initial_completion_budget_tokens": 16384,
+          "max_repair_rounds": 2,
+          "model_id": "gemini/gemini-4-argon",
+          "prompt_contract_version": "2026-08-09-v2-scoring-contract",
+          "request_timeout_seconds": 300,
+          "thinking": {
+            "mode": "provider_default"
+          },
+          "tool_choice_mode": "forced"
+        }
+      },
+      "provider_id": "gemini/gemini-4-argon",
+      "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": "forced"
+    },
     "glm-5.2": {
       "answer_contract": "json",
       "evidence": {
@@ -1493,6 +1550,63 @@
       "shared_completion_budget_tokens": 16384,
       "tool_choice": "forced"
     },
+    "muse-spark-1.3": {
+      "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": "a867a064b1fa887213b8d3493112bea16c1967a97b639536e6a4c39695a801fe",
+        "kind": "run_state",
+        "rows_match": true,
+        "run": "argon-runs/musespark",
+        "source_predictions_sha256": "1e4c81a1c69b4be3dec92d007a2c3d3c22009bf7d56df29a2d2349a5be4887d3",
+        "source_rows_sha256": "a867a064b1fa887213b8d3493112bea16c1967a97b639536e6a4c39695a801fe",
+        "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-spark-1.3",
+          "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-spark-1.3",
+      "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
+    },
     "muse-wonder": {
       "answer_contract": "json",
       "evidence": {
@@ -1642,7 +1756,7 @@
       "tool_choice": null
     }
   },
-  "registry_commit": "3ef45909db7367abdffd08379babde5bcf5c549e",
+  "registry_commit": "236134f5c3b4400a14ae7de24604c988f6f602eb",
   "sources": [
     "policybench.config.MODELS",
     "policybench.model_cards",

diff --git a/app/src/modelMeta.ts b/app/src/modelMeta.ts
@@ -31,6 +31,7 @@ export const MODEL_ORDER = [
   "gpt-5.5",
   "gpt-5.4-mini",
   "gpt-5.4-nano",
+  "gemini-4-argon",
   "gemini-3.1-pro-preview",
   "gemini-3.6-flash",
   "gemini-3.8-flash",
@@ -53,6 +54,7 @@ export const MODEL_ORDER = [
   "qwen3.8-max",
   "qwen-3.7-max",
   "muse-wonder",
+  "muse-spark-1.3",
 ] as const;
 
 export const MODEL_LABELS: Record<string, string> = {
@@ -105,6 +107,8 @@ export const MODEL_LABELS: Record<string, string> = {
   "qwen3.8-max": "Qwen 3.8 Max",
   inkling: "Inkling",
   "muse-wonder": "Muse Wonder",
+  "muse-spark-1.3": "Muse Spark 1.3",
+  "gemini-4-argon": "Gemini 4 Argon",
 };
 
 export type ProviderKey =
@@ -227,7 +231,7 @@ const FRONTIER_MODEL_GROUPS = [
   ["claude-fable-5.1", "claude-fable-5", "claude-opus-5", "claude-opus-4.8"],
   ["gpt-6-astra", "gpt-5.6-sol", "gpt-5.5"],
   ["grok-4.5", "grok-4.3"],
-  ["gemini-3.1-pro-preview"],
+  ["gemini-4-argon", "gemini-3.1-pro-preview"],
   ["deepseek-v4-pro-0813", "deepseek-v4-pro"],
   ["kimi-k3", "kimi-k2.6"],
   ["glm-5.3", "glm-5.2"],

diff --git a/app/src/notes/2026-10-14-gemini-4-argon-joins-the-board.json b/app/src/notes/2026-10-14-gemini-4-argon-joins-the-board.json
@@ -0,0 +1,65 @@
+{
+  "slug": "2026-10-14-gemini-4-argon-joins-the-board",
+  "date": "2026-10-14",
+  "title": "Google's Gemini 4 Argon and Meta's Muse Spark 1.3 join the board",
+  "paragraphs": [
+    "Google's Gemini 4 Argon joined the board on {releaseDate}, the day after Google opened it to the public Gemini API, and Meta's Muse Spark 1.3 joined with it, bringing the board to {nModels} models. Gemini 4 Argon gets {argonExact}% of answers within $1, weighted by household impact, #{argonRank} of {nModels} and the highest of any Gemini model; Gemini 3.8 Flash, the previous best, gets {flashExact}% (#{flashRank}). Muse Spark 1.3 gets {sparkExact}% (#{sparkRank}), behind Meta's Muse Wonder ({museExact}%, #{museRank}), but on the bounded score, which gives partial credit for amounts close to the reference, it ranks #{sparkBoundedRank}. GPT-6 Sol still leads on answers within $1 at {solExact}%, ahead of Claude Opus 5.5 ({opusExact}%) and GPT-5.6 Sol ({sol56Exact}%).",
+    "Google announced Gemini 4 Argon on September 30 for a group of partners and released it on the Gemini API on October 13. PolicyBench runs it there, as it does every Gemini model. The API accepts a forced answer tool, so the row answers through the board's tool and reasons at Google's default setting. A forced tool call suppresses some models' reasoning, as it does Claude Opus 5's, so PolicyBench fixed a test before looking at the results and ran all {households} households both ways. With the forced tool Gemini 4 Argon returned reasoning tokens on {forcedReasoning} of them, and on {boardReasoning} in the board's own run; with an automatic tool choice, on {autoReasoning}. The test found no significant reduction, so the board keeps the forced tool and publishes no second row. At Google's introductory price of $2 per million input tokens and $10 per million output tokens, which doubles on December 1, the {households} households cost ${argonCost}.",
+    "Muse Spark 1.3, which Meta released on September 2, is Meta's second model on the board. PolicyBench reaches it through OpenRouter, whose one endpoint for it is Meta's own API, as it does Muse Wonder. That API rejects a forced tool call, so the row answers as a JSON object and reasons at Meta's default setting.",
+    "Nothing else on the board moves. The newest policyengine-us release, {checkedEngine}, gives the same value as {engineVersion} for every output, so the references stay where they were, and every model is still scored on {scoredOutputs} of its {totalOutputs} requested outputs. Each of the {incumbents} earlier models keeps its score."
+  ],
+  "facts": {
+    "releaseDate": "2026-10-14",
+    "nModels": 50,
+    "argonExact": 88.7,
+    "argonRank": 12,
+    "flashExact": 85.6,
+    "flashRank": 21,
+    "sparkExact": 87.2,
+    "sparkRank": 18,
+    "museExact": 90.1,
+    "museRank": 10,
+    "sparkBoundedRank": 6,
+    "solExact": 95.2,
+    "opusExact": 94.1,
+    "sol56Exact": 93.1,
+    "households": 100,
+    "forcedReasoning": 80,
+    "boardReasoning": 85,
+    "autoReasoning": 82,
+    "argonCost": "2.79",
+    "checkedEngine": "2.42.4",
+    "engineVersion": "2.41.0",
+    "scoredOutputs": 1935,
+    "totalOutputs": 1984,
+    "incumbents": 48
+  },
+  "data": [
+    {
+      "label": "Dashboard data release",
+      "href": "https://github.com/PolicyEngine/policybench/releases/tag/dashboard-data-20261014"
+    },
+    {
+      "label": "Release design note",
+      "href": "https://github.com/PolicyEngine/policybench/blob/main/docs/argon/design.md"
+    },
+    {
+      "label": "Publication check against policyengine-us 2.42.4",
+      "href": "https://github.com/PolicyEngine/policybench/blob/main/reference_audit/2026-10-12-engine-upgrade/verification/publication_check_2_42_4.csv"
+    },
+    {
+      "label": "Gemini 4 Argon model page",
+      "href": "/model/gemini-4-argon"
+    },
+    {
+      "label": "Muse Spark 1.3 model page",
+      "href": "/model/muse-spark-1.3"
+    },
+    {
+      "label": "Previous release note (October 13)",
+      "href": "/notes/2026-10-13-muse-wonder-joins-the-board"
+    }
+  ],
+  "boardSnapshot": "2026-10-14",
+  "release": "dashboard-data-20261014"
+}
diff --git a/app/src/notes/index.ts b/app/src/notes/index.ts
@@ -1,3 +1,4 @@
+import gemini4ArgonJoinsTheBoard from "./2026-10-14-gemini-4-argon-joins-the-board.json";
 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";
@@ -31,6 +32,7 @@ export type PolicyBenchNote = {
 };
 
 export const notes = [
+  gemini4ArgonJoinsTheBoard,
   museWonderJoinsTheBoard,
   claudeHaiku55JoinsTheBoard,
   stopsScoringEightTaxOutputs,

diff --git a/app/src/paperSnapshot.json b/app/src/paperSnapshot.json
@@ -1,6 +1,6 @@
 {
-  "pdfVersion": "cee26d0956da",
+  "pdfVersion": "d0b236367fb3",
   "responseWindow": "June 12 and October 13, 2026",
-  "snapshotDate": "2026-10-13",
-  "webVersion": "320e58e44470"
+  "snapshotDate": "2026-10-14",
+  "webVersion": "533a0cb79b6a"
 }
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 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).
+    // Release 20261014: the counts tests/test_disclosures.py recomputes
+    // from the frozen annotations and payload (50 models; Gemini 4 Argon's
+    // and Muse Spark 1.3's rows join on release 20261013's references and
+    // exclusions).
     expect(summarizeAuditUniverse(us)).toEqual({
-      annotatedRowCount: 8_503,
-      legacyThresholdRowCount: 8_503,
-      exactMissCount: 8_499,
-      annotatedExactMissCount: 8_499,
+      annotatedRowCount: 8_765,
+      legacyThresholdRowCount: 8_765,
+      exactMissCount: 8_761,
+      annotatedExactMissCount: 8_761,
       annotatedExactHitCount: 4,
-      unannotatedBelowFullBoundedScoreCount: 2_218,
+      unannotatedBelowFullBoundedScoreCount: 2_329,
     });
   });
 });
diff --git a/app/tests/modelMeta.test.ts b/app/tests/modelMeta.test.ts
@@ -30,6 +30,8 @@ describe("model metadata", () => {
     ["grok-4.7", "Grok 4.7", "xai"],
     ["deepseek-v4.1-flash", "DeepSeek V4.1 Flash", "deepseek"],
     ["muse-wonder", "Muse Wonder", "meta"],
+    ["muse-spark-1.3", "Muse Spark 1.3", "meta"],
+    ["gemini-4-argon", "Gemini 4 Argon", "google"],
   ] as const)("maps %s to its label and provider", (model, label, provider) => {
     expect(MODEL_LABELS[model]).toBe(label);
     expect(getProviderForModel(model)).toBe(provider);
@@ -80,6 +82,21 @@ describe("model metadata", () => {
     ]);
   });
 
+  test("shows Gemini 4 Argon as Google's flagship, and 3.1 Pro without it", () => {
+    expect(
+      frontierModelsFor(["gemini-4-argon", "gemini-3.1-pro-preview"]),
+    ).toEqual(["gemini-4-argon"]);
+    expect(frontierModelsFor(["gemini-3.1-pro-preview"])).toEqual([
+      "gemini-3.1-pro-preview",
+    ]);
+  });
+
+  test("keeps Muse Wonder as Meta's flagship over Muse Spark 1.3", () => {
+    expect(frontierModelsFor(["muse-spark-1.3", "muse-wonder"])).toEqual([
+      "muse-wonder",
+    ]);
+  });
+
   test("gives every represented provider a frontier model", () => {
     const models = [
       "claude-fable-5",
@@ -123,7 +140,6 @@ describe("MODEL_RELEASE_DATES", () => {
   });
 });
 
-
 describe("shortModelLabel", () => {
   test("drops a parenthetical qualifier before keeping the last two words", () => {
     expect(shortModelLabel("GLM-5.3-Flash (preview)")).toBe("GLM-5.3-Flash");
@@ -136,8 +152,12 @@ describe("shortModelLabel", () => {
 describe("MODEL_PROVENANCE_NOTES", () => {
   test("the GLM-5.3-Flash preview row records the listing it was queried under", () => {
     expect(MODEL_LABELS["ox-alpha"]).toBe("GLM-5.3-Flash (preview)");
-    expect(MODEL_PROVENANCE_NOTES["ox-alpha"]).toContain('Queried as "Ox Alpha"');
+    expect(MODEL_PROVENANCE_NOTES["ox-alpha"]).toContain(
+      'Queried as "Ox Alpha"',
+    );
     expect(MODEL_PROVENANCE_NOTES["ox-alpha"]).toContain("2026-08-20");
-    expect(Object.keys(MODEL_PROVENANCE_NOTES).every((id) => id in MODEL_LABELS)).toBe(true);
+    expect(
+      Object.keys(MODEL_PROVENANCE_NOTES).every((id) => id in MODEL_LABELS),
+    ).toBe(true);
   });
 });