AI model scores ≥ 90% on FrontierMath Benchmark before 2027?

Traders currently price "Yes" at 90% for "AI model scores ≥ 90% on FrontierMath Benchmark before 2027?". The market has traded $117K in total volume and resolves Dec 31, 2026. Rapid progress on FrontierMath, Epoch AI’s benchmark of unpublished research-level math problems, underpins the 89.5% market-implied probability that an AI model reaches ≥90% before 2027. Top systems such as Claude Fable 5 and GPT-5.6 variants have climbed from under 3% in late 2024 to 82–88% on the corrected v2 dataset released in June 2026, driven by scaled training, advanced reasoning chains, and Python tool integration. Ongoing releases from OpenAI, Anthropic, and DeepMind, combined with historical acceleration on hard math benchmarks, sustain trader consensus that the remaining gap will close within the timeframe despite error bars and potential timeline slips.

Odds sourced live from Polymarket, refreshed through the day. Prediction-market participation carries risk and may be restricted in your jurisdiction. Informational only — not investment advice.