Best AI model on LiveBench (Mathematics) at end of October: Anthropic leads at 79%, OpenAI at 23% on $51.8K traded

Anthropic leads, with differing probability snapshots

Anthropic is the favorite to finish October with the best AI model on LiveBench’s Mathematics leaderboard, according to Polymarket. The headline snapshot published Oct 15, 2026, at 6:09 PM ET puts Anthropic at 79% and OpenAI at 23%, with $51.8K traded.

The supplied Polymarket account gives different probabilities: 74.5% for Anthropic, 18.5% for OpenAI and 5.9% for Google. The material does not establish why those figures differ from the headline snapshot, so they should be treated as separate readings rather than a documented price move.

Mathematics scores drive the contest

According to Polymarket, Anthropic’s Claude Opus 5.5 and Fable 5.1 recorded the highest Mathematics scores in the October 14 leaderboard refresh, beating OpenAI’s entries on verifiable quantitative reasoning tasks. OpenAI’s o-series and GPT-6 variants remain competitive overall, the account says, but lag in this particular category. It attributes Google’s smaller probability to weaker recent Mathematics results.

Polymarket’s account links traders’ confidence in Anthropic to its sustained emphasis on structured problem-solving benchmarks and an absence of comparable OpenAI or Google releases in the final two weeks. That is the account’s explanation of market sentiment, not a guarantee that the leaderboard will remain unchanged: it also identifies a late model update as a potential source of disruption before the end-of-October resolution.

The accompanying editorial summary describes the preceding 15 minutes as lacking a significant new platform headline. It says this fresh contract consequently led coverage, shifting into a technology category different from the previous three items.

What it means for prediction markets

• The category matters: This contract concerns LiveBench Mathematics at the end of October 2026, not overall AI-model performance.
• The snapshots need separating: The headline’s 79% and 23% differ from the supplied account’s 74.5% and 18.5%. Without further timing information, the discrepancy does not establish a trading trend.
• Late releases remain relevant: The October 14 results support the reported Anthropic lead, but a subsequent model update could change the leaderboard before resolution.
• Volume is context, not confirmation: The headline reports $51.8K traded; that figure does not settle which company will ultimately lead.