Meta's Odds of a Top-Ranked AI Model Hit 14%, but Muse Is 9th

---

Meta Is Not #1 on LM Arena, but the Market Gives It a 14% Chance of a Top-Ranked Model This Year

Meta does not have a top-ranked AI model right now. On the LM Arena text overall leaderboard snapshot dated Oct. 8, 2026, Google's gemini-4-argon-high ranked #1 with a preliminary score of 1525, and Anthropic's Claude models came next. Meta's best entry, muse-spark-1.3-max, ranked #9 at 1494, 31 points behind the leader.

On Polymarket, Meta's contract in "Which companies will have a top-ranked AI model this year?" rose from 8% to 14% over three days, a gain of about 6 percentage points. That price stood at 14% on Oct. 16. It is a market-implied probability: a trading price set by people buying and selling the contract. It is not a poll or a model forecast.

Live market odds

2 venues

Which companies will have a top-ranked AI model this year?

Each outcome's chance, from what traders are paying right now on Kalshi and Polymarket.

Will Meta have a #1 AI model by December 31, 2026? [Yes]In this story

8%

KalshiBest price11.11x9¢Polymarket10.00x10¢

Nvidia

2%

Kalshi33.33x3¢

Will xAI have a #1 AI model by December 31, 2026? [Yes]

8%

PolymarketBest price12.50x8¢Kalshi10.00x10¢

Will Mistral have a #1 AI model by December 31, 2026? [Yes]

3%

KalshiBest price33.33x3¢Polymarket33.33x3¢

New to Kalshi? Get a $35 trading bonus

Code PRED35. New users only. Eligibility restrictions and terms apply.

Claim offer

Compare every venueCents per $1 payout, before fees

The leaderboard rank is the main evidence against Meta. Meta itself points to two newer releases.

---

Muse Wonder and Muse Spark Open Are Meta's Strongest Claims

Meta introduced Muse Wonder and Muse Wonder Full on Oct. 12. It calls them its most capable models and says they deliver "state-of-the-art performance on personal and professional agentic tasks." Meta reports state-of-the-art results on benchmarks for agents, memory and coding. These figures come from Meta's own evaluation report, and the models had not been placed on the general text leaderboard in the Oct. 8 snapshot. Muse Wonder is available through the Meta Model API and Muse Code. Muse Wonder Full is limited to Muse Code subscribers.

On Oct. 15, Meta released the weights for Muse Spark Open, a model with 816B total parameters, 42B of them active, and a 1M-token context window. Meta calls it "the most capable US open-weight model to date." That claim is Meta's own and is limited to US models. None of the sources here rank it as the best open-weight model worldwide.

Both releases came during the period when the contract rose, but we could not establish that either one caused the move.

---

What the Move to 14% Does and Doesn't Say About Meta's Odds

1D7D30D

The 8% starting point is the period low, so the contract has gained steadily since then. A 14% price still implies roughly an 86% chance that Meta does not finish the year with a top-ranked model under the market's terms.

Meta sits ninth on the general board. One reading of the price is that traders are backing a long shot that depends on Meta's strength in specific areas. Meta's Arena organization page shows strong standings on narrower boards such as MultiChallenge and MCP Atlas. That evidence does not point to Meta taking the overall #1 spot.

How the contract settles depends on the market's own definition of "top-ranked," including which leaderboard or category counts and when it is measured. The sources reviewed here do not spell out those terms, so readers should check the contract rules on the market page for which companies will have a top-ranked AI model this year before reading the 14% as a verdict on Meta's models. The market resolves on Jan. 2, 2027.

Two things would change the picture: Muse Wonder climbing the LM Arena text board toward Google's 1525, or a ranking the market's rules count placing Meta first.