november ai model odds: Google 75.5%, Anthropic 23.5% After Fable 5.2

The november ai model odds on Polymarket still favor Google by a wide margin: at our gamma pull around 06:01 UTC on November 10, "Google has the best AI model at the end of November" traded near 75.5% Yes, with Anthropic at about 23.5% and OpenAI near 2.1%. That is the morning after Anthropic released Claude Fable 5.2, its new flagship, so traders are now watching whether it can climb the chat arena in three weeks.

What actually happened

Two launches on November 9 put this market back in focus. Anthropic released Claude Fable 5.2 and Mythos 5.2, claiming leading scores on several coding and reasoning tests in its own table. Meta released a small model, Muse Ember, that does not target the top of the arena. Neither had a settled Arena rank at our check, because new models need days or weeks of votes before their ratings stabilize.

The current leader is Google. BenchLeader, which republishes Arena data under a CC BY license, listed Gemini 4 Carbon (high) at 1527 and Gemini 4 Argon (high) at 1522 as of November 10, followed by Meta's Muse Wonder at 1512 and Claude Opus 5.5 at 1506. Claude Fable 5 and Fable 5.1 sat at 1504 and 1503. We covered those rivals in muse wonder and claude opus 5.5. That gap of more than 20 points between Gemini 4 Carbon and the best Claude model is what the 75.5% price reflects.

There is a detail to keep in mind: BenchLeader shows the style-controlled board, while this market uses the board with style control off. Rankings on the two views are often close but not identical.

What Polymarket thinks

"Which company has the best AI model end of November?" is one of the busiest AI markets on the site, with about $1.10M of total volume and around $141K traded in the last 24 hours at our pull. Here are the november ai model odds:

• Google: 75.5% Yes (bid 75, ask 76)
• Anthropic: 23.5% Yes (bid 23, ask 24)
• OpenAI: 2.1% Yes
• Meta, Alibaba, Z.ai, SpaceXAI: 0.15% Yes each

The book is tight on the two leaders, which means real money sits on both sides. A sister market that uses the style-control-on board showed almost the same split, Google about 72.5% and Anthropic about 23.5%. We tracked the longer view in year-end ai model odds.

What could move it

For Anthropic to win, a Claude model has to be ranked first on the style-off board at the check time. Ties on rank are broken by the underlying Arena score, then alphabetically by company name, which favors "Anthropic" over "Google" only in an exact dead heat. Fable 5.2 is the obvious candidate, but Anthropic's own benchmark wins do not guarantee arena wins, since arena votes measure which answer people prefer in chat, not coding or agent tasks.

For Google, the risk is mostly a new competitor. A surprise release from any lab before November 30 could change the order, which is why the "other company" lines are not exactly zero. Models marked "AutoEval" on the board at check time do not count.

How the arena scores work

Arena ratings come from people comparing two anonymous answers to the same prompt and picking the better one. Millions of those votes are turned into an Elo-style score. BenchLeader notes that a 100-point gap means the higher model wins about 64% of head-to-head votes, so a gap of around 20 points is real but not huge. It also warns that the voter pool is self-selected and that providers can test unreleased variants anonymously before launch.

What we could not verify

We could not load the live arena.ai board directly from our tools, so the rankings above come from BenchLeader's republished data, which lags the live board by a day or two. Fable 5.2 had no published Arena score at our check.

What to do as a reader

Treat the november ai model odds as a read on how traders weigh one new launch against an established leader. Check the live Polymarket market and the arena.ai leaderboard with style control off before relying on any number here. Nothing here is investment advice, and arena ratings for new models can swing for days after launch.