{"data":{"event":{"id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","slug":"frontiermath-erdos-astra-prerelease","title":"Pre-release GPT-6 Astra solves five Bloom-curated Erdős problems in Lean","short_summary":"On 1 September 2026 Epoch AI reported that a pre-release GPT-6 Astra system was the only model to score on FrontierMath Erdős, a Lean benchmark of 68 significant unsolved Erdős problems. It solved problems 74 and 126 under the official $300 protocol, and problems 1, 548, and 571 in extra attempts.","full_description":"Thomas Bloom selected 68 Erdős problems that were open as of August 2026. Epoch ran GPT-6 Astra (pre-release), GPT-5.6 Sol, GPT-5.5, and Claude Fable 5/5.1 at $300 and 72 hours per problem. Only Astra scored: a counterexample to problem 74 and a proof of problem 126 (3%). Additional, non-protocol attempts with larger budgets also yielded a disproof of problem 1 and proofs of 548 and 571. Those extras are not a benchmark score. This is recorded as realized Lean-checked mathematics on named Erdős problems, not as a 3% exam result. It is distinct from the May unit-distance disproof (problem 90) and from the August ten-advances cluster.","ledger_type":"benefit","primary_domain_id":"23d935bb-b857-4b2b-9471-d568bb848a66","event_status":"provisional","event_date":"2026-09-01T00:00:00.000Z","discovery_date":"2026-09-01T00:00:00.000Z","first_published_date":"2026-09-01T00:00:00.000Z","last_reviewed_date":"2026-09-01T00:00:00.000Z","geographic_scope":"International","affected_population":"Combinatorics and Erdős-problem research community","base_impact_tier":2,"base_score":"3.00","attribution_multiplier":"0.8500","evidence_multiplier":"0.5000","realization_multiplier":"0.5500","durability_multiplier":"0.6500","current_event_score":"0.455800","confidence_level":"medium","score_explanation":"Lean-checked solutions to Bloom-curated Erdős problems are notable mathematics (tier 2), not a field-defining cluster. Attribution is to pre-release Astra (0.85). Evidence is Epoch’s public protocol and paper (0.50). Realization is verified Lean, with extra solves outside the official score (0.55).","methodology_version_id":"d7881163-fb23-4e71-8625-bac1a8662c0f","original_methodology_version_id":"d7881163-fb23-4e71-8625-bac1a8662c0f","published_at":"2026-09-01T00:00:00.000Z","created_at":"2026-09-13T22:56:05.431Z","updated_at":"2026-09-13T22:56:05.431Z","flags":[],"domain_name":"Mathematics","domain_slug":"mathematics","methodology_version":"0.1"},"contributions":[{"id":"e36e45c9-01d6-43f9-a283-7ca7035f59de","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","model_id":"0f0d18fc-d3f1-4551-9eaf-a13f1f93b1a5","role_description":"Pre-release GPT-6 Astra run by Epoch on FrontierMath Erdős; the only scored model in the official protocol.","attribution_multiplier":"0.8500","credit_share":"1.0000","contribution_score":"0.455800","attribution_rationale":"Epoch names a pre-release GPT-6 Astra as the solver. That is the named Astra line, not the August internal ten-proofs system or the September Navier–Stokes search model.","attribution_confidence":"medium","first_used_date":"2026-09-01T00:00:00.000Z","model_version_if_known":null,"review_status":"approved","created_at":"2026-09-13T22:56:05.431Z","model_slug":"gpt-6-astra","model_name":"GPT-6 Astra","identity_class":"named_release","is_internal":false,"family_name":"GPT-6","family_slug":"openai-gpt-6","organization_name":"OpenAI","organization_slug":"openai"}],"sources":[{"id":"2b2e05e8-7b29-4532-b77b-bb91d9e4c7a5","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","url":"https://epoch.ai/latest/announcing-frontiermath-erdos","canonical_url":"https://epoch.ai/latest/announcing-frontiermath-erdos","source_type":"other","publisher":"Epoch AI","author":null,"publication_date":"2026-09-01T00:00:00.000Z","retrieved_at":"2026-09-13T22:56:05.431Z","title":"Announcing FrontierMath Erdős","excerpt":null,"content_hash":null,"source_reliability_class":"medium","is_primary_source":true,"is_independent":true,"is_peer_reviewed":false,"archived_url":null,"created_at":"2026-09-13T22:56:05.431Z"},{"id":"b396d6f2-017e-499b-a91d-eb57c18b559a","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","url":"https://epoch.ai/benchmarks/frontiermath-erdos","canonical_url":"https://epoch.ai/benchmarks/frontiermath-erdos","source_type":"other","publisher":"Epoch AI","author":null,"publication_date":"2026-09-01T00:00:00.000Z","retrieved_at":"2026-09-13T22:56:05.431Z","title":"FrontierMath Erdős benchmark","excerpt":null,"content_hash":null,"source_reliability_class":"medium","is_primary_source":false,"is_independent":true,"is_peer_reviewed":false,"archived_url":null,"created_at":"2026-09-13T22:56:05.431Z"}],"claims":[{"id":"554001f7-177e-4071-ba53-647f4f85e7d4","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","claim_text":"Under Epoch’s $300 protocol, pre-release GPT-6 Astra produced Lean-verified solutions to Erdős problems 74 and 126, for a 3% score; Sol and Fable scored 0%.","claim_type":"outcome","claim_status":"supported","confidence_score":null,"created_at":"2026-09-13T22:56:05.431Z","updated_at":"2026-09-13T22:56:05.431Z"},{"id":"89518e21-8e59-40cd-9215-108087dc7ef0","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","claim_text":"Additional larger-budget attempts also yielded Lean solutions to problems 1, 548, and 571; those are not part of the official 3% score.","claim_type":"outcome","claim_status":"supported","confidence_score":null,"created_at":"2026-09-13T22:56:05.431Z","updated_at":"2026-09-13T22:56:05.431Z"}],"revisions":[{"id":"820bdc88-5b01-40c4-b272-bf3607f88e13","event_id":"0a1563be-aacc-4a01-94b9-e0a4082cf18d","model_id":null,"previous_score":"0.000000","new_score":"0.455800","previous_factors":{},"new_factors":{"evidence":0.5,"base_score":3,"durability":0.65,"attribution":0.85,"realization":0.55},"change_reason":"Imported events under methodology 0.1.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-09-13T22:56:05.431Z"}],"secondary":[{"name":"Computer Science","slug":"computer-science"}]},"methodology_version":"0.1"}