AI football prediction leaderboard

AI Football Arena

Frontier AI models predict real football matches on the same terms. Each pick locks before kickoff, then the result settles it. Arena Score ranks the field across every competition.

Same games onlyRestrict the final 13 models to the cohort games every model completed.

13 models7,533 scored predictions

Full record leaderboard for the final 13 AI models

1	Claude Opus 5 Anthropic	67.1	+1.7%53% vs 51.3% mkt	13%	+6.7%−2% vs −8.7% mkt	555
2	GLM-5.2 Z.ai	66.2	+1.3%53% vs 51.7% mkt	12%	+8.6%0.7% vs −7.9% mkt	578
3	Grok 4.5 xAI	62.5	+1.1%53% vs 51.9% mkt	12%	+5.9%−1.5% vs −7.4% mkt	577
4	GPT-5.6 Sol OpenAI	60.4	+0.3%52% vs 51.7% mkt	12%	+5.6%−2.3% vs −7.9% mkt	578
5	Nemotron 3 Ultra NVIDIA	59.5	−0.6%51% vs 51.6% mkt	11%	+9.0%0.9% vs −8.1% mkt	575
6	Kimi K3 Moonshot	58.2	+0.9%52% vs 51.1% mkt	11%	+4.4%−4.3% vs −8.7% mkt	563
7	Qwen3.7 Plus Alibaba	56.4	−0.7%51% vs 51.7% mkt	13%	+2.4%−5.3% vs −7.7% mkt	579
8	Mistral Large 3 Mistral	56.1	−0.9%53% vs 53.9% mkt	12%	+6.4%0.2% vs −6.2% mkt	608
9	MiniMax M3 MiniMax	54.9	−0.7%51% vs 51.7% mkt	12%	+3.0%−4.7% vs −7.7% mkt	579
10	Gemini 3.7 Flash Google	53.2	−1.3%50% vs 51.3% mkt	13%	0.0%−7.8% vs −7.8% mkt	478
11	Muse Spark 1.2 Meta	51.4	−1.6%50% vs 51.6% mkt	13%	−0.4%−7.4% vs −7.0% mkt	498
12	DeepSeek V4 Pro DeepSeek	50.6	−1.7%52% vs 53.7% mkt	12%	+2.5%−3.5% vs −6.0% mkt	682
13	MiMo v2.5-Pro Xiaomi	49.9	−0.7%53% vs 53.7% mkt	10%	+3.5%−2.5% vs −6.0% mkt	683

1Claude Opus 5

Anthropic67.1Arena score

Acc Δ

+1.7%

Exact

13%

ROI Δ

+6.7%

555

2GLM-5.2

Z.ai66.2Arena score

Acc Δ

+1.3%

Exact

12%

ROI Δ

+8.6%

578

3Grok 4.5

xAI62.5Arena score

Acc Δ

+1.1%

Exact

12%

ROI Δ

+5.9%

577

4GPT-5.6 Sol

OpenAI60.4Arena score

Acc Δ

+0.3%

Exact

12%

ROI Δ

+5.6%

578

5Nemotron 3 Ultra

NVIDIA59.5Arena score

Acc Δ

−0.6%

Exact

11%

ROI Δ

+9.0%

575

6Kimi K3

Moonshot58.2Arena score

Acc Δ

+0.9%

Exact

11%

ROI Δ

+4.4%

563

7Qwen3.7 Plus

Alibaba56.4Arena score

Acc Δ

−0.7%

Exact

13%

ROI Δ

+2.4%

579

8Mistral Large 3

Mistral56.1Arena score

Acc Δ

−0.9%

Exact

12%

ROI Δ

+6.4%

608

9MiniMax M3

MiniMax54.9Arena score

Acc Δ

−0.7%

Exact

12%

ROI Δ

+3.0%

579

10Gemini 3.7 Flash

Google53.2Arena score

Acc Δ

−1.3%

Exact

13%

ROI Δ

0.0%

478

11Muse Spark 1.2

Meta51.4Arena score

Acc Δ

−1.6%

Exact

13%

ROI Δ

−0.4%

498

12DeepSeek V4 Pro

DeepSeek50.6Arena score

Acc Δ

−1.7%

Exact

12%

ROI Δ

+2.5%

682

13MiMo v2.5-Pro

Xiaomi49.9Arena score

Acc Δ

−0.7%

Exact

10%

ROI Δ

+3.5%

683

This is the full record for the final 13-model field: every fixture the site has graded for each of them. Models joined at different times, so they hold different fixtures, which is why the columns compare each one with the market on its own games. Superseded models remain in the full-history leaderboards and model directory. How the three boards differ →

13 models397 / 1,000 common games

Same-games AI model leaderboard

1	Claude Opus 5 Anthropic	69.6	+1.4%52% vs 50.6% mkt	14%	+8.7%−0.7% vs −9.4% mkt	397
2	Mistral Large 3 Mistral	62.4	+0.4%51% vs 50.6% mkt	12%	+8.1%−1.3% vs −9.4% mkt	397
3	GLM-5.2 Z.ai	62.3	+0.4%51% vs 50.6% mkt	12%	+8.0%−1.4% vs −9.4% mkt	397
4	GPT-5.6 Sol OpenAI	60.7	+0.4%51% vs 50.6% mkt	12%	+6.5%−2.9% vs −9.4% mkt	397
5	DeepSeek V4 Pro DeepSeek	58.3	−0.6%50% vs 50.6% mkt	13%	+4.6%−4.8% vs −9.4% mkt	397
6	Grok 4.5 xAI	57.7	+0.4%51% vs 50.6% mkt	12%	+3.6%−5.8% vs −9.4% mkt	397
7	Gemini 3.7 Flash Google	57.6	−0.6%50% vs 50.6% mkt	14%	+1.9%−7.5% vs −9.4% mkt	397
8	Kimi K3 Moonshot	56.1	+0.4%51% vs 50.6% mkt	11%	+4.1%−5.3% vs −9.4% mkt	397
9	Nemotron 3 Ultra NVIDIA	55.7	−1.6%49% vs 50.6% mkt	11%	+8.4%−1% vs −9.4% mkt	397
10	Qwen3.7 Plus Alibaba	53.0	−1.6%49% vs 50.6% mkt	13%	+1.8%−7.6% vs −9.4% mkt	397
11	MiniMax M3 MiniMax	52.9	−0.6%50% vs 50.6% mkt	11%	+3.4%−6% vs −9.4% mkt	397
12	Muse Spark 1.2 Meta	49.9	−2.6%48% vs 50.6% mkt	14%	−0.8%−10.2% vs −9.4% mkt	397
13	MiMo v2.5-Pro Xiaomi	49.8	−1.6%49% vs 50.6% mkt	11%	+2.8%−6.6% vs −9.4% mkt	397

1Claude Opus 5

Anthropic69.6Arena score

Acc Δ

+1.4%

Exact

14%

ROI Δ

+8.7%

397

2Mistral Large 3

Mistral62.4Arena score

Acc Δ

+0.4%

Exact

12%

ROI Δ

+8.1%

397

3GLM-5.2

Z.ai62.3Arena score

Acc Δ

+0.4%

Exact

12%

ROI Δ

+8.0%

397

4GPT-5.6 Sol

OpenAI60.7Arena score

Acc Δ

+0.4%

Exact

12%

ROI Δ

+6.5%

397

5DeepSeek V4 Pro

DeepSeek58.3Arena score

Acc Δ

−0.6%

Exact

13%

ROI Δ

+4.6%

397

6Grok 4.5

xAI57.7Arena score

Acc Δ

+0.4%

Exact

12%

ROI Δ

+3.6%

397

7Gemini 3.7 Flash

Google57.6Arena score

Acc Δ

−0.6%

Exact

14%

ROI Δ

+1.9%

397

8Kimi K3

Moonshot56.1Arena score

Acc Δ

+0.4%

Exact

11%

ROI Δ

+4.1%

397

9Nemotron 3 Ultra

NVIDIA55.7Arena score

Acc Δ

−1.6%

Exact

11%

ROI Δ

+8.4%

397

10Qwen3.7 Plus

Alibaba53.0Arena score

Acc Δ

−1.6%

Exact

13%

ROI Δ

+1.8%

397

11MiniMax M3

MiniMax52.9Arena score

Acc Δ

−0.6%

Exact

11%

ROI Δ

+3.4%

397

12Muse Spark 1.2

Meta49.9Arena score

Acc Δ

−2.6%

Exact

14%

ROI Δ

−0.8%

397

13MiMo v2.5-Pro

Xiaomi49.8Arena score

Acc Δ

−1.6%

Exact

11%

ROI Δ

+2.8%

397

This view uses the fixed 13-model study roster and the 397 cohort fixtures with a result, an eligible pre-match market snapshot and a grade from every model. The 1,000 fixtures were fixed in advance; incomplete fixtures stay outside this board. Study design →

Arena Score blends 90-minute accuracy, exact score and betting ROI into one number, measured against the market on the same games. 50 is the market baseline. Above it, a model beat the market. Below it, the market beat the model. Acc Δ and ROI Δ do the same job one metric at a time: each model against a bettor who always backs the shortest pre-match price, on the fixtures that model played.

Leaderboards by skill Browse all fixtures

Upcoming fixtures

See all fixtures →

Sep 30 · 23:30 UTCMajor League SoccerNew York Red BullsvsSt. Louis CityPredictions made Oct 1 · 16:00 UTCUEFA Nations LeagueAzerbaijanvsLiechtensteinPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueDenmarkvsPortugalPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueGermanyvsSerbiaPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueWalesvsNorwayPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueRep. Of IrelandvsAustriaPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueMaltavsGibraltarPredictions made Oct 1 · 18:45 UTCUEFA Nations LeagueIsraelvsKosovoPredictions made

Questions & answers

Can AI models predict football matches?

Nobody knows yet, which is why this site runs the experiment. Frontier LLMs from Claude, GPT, Gemini, Grok, DeepSeek, Mistral, GLM, Kimi and others all take the same matches on the same terms: one prompt, the same research tools, a fixed temperature. Picks lock before kickoff, and the 90-minute result decides them. What the record shows so far →

Which AI is best at predicting football?

The main leaderboard ranks every active model by Arena Score, a composite of 90-minute accuracy, exact score, betting ROI and probability calibration, all measured against the market favourite on the same fixtures. Whoever leads sits at the top of that board. The models directory lists the whole field.

Can AI beat the bookies at football?

That is what the betting ROI leaderboard tracks: each model staked flat against the market favourite on its own fixtures, ranked on the gap between them. So far none has beaten the market consistently across a full competition. Beating it for a week is easy and means nothing. The blog follows the chase in more detail. This is a benchmark, and nothing on it is betting advice.

How are AI football predictions scored?

A model returns a home/draw/away probability distribution plus its single most likely exact score. Its pick is whichever probability is highest. A hit means that pick matched the 90-minute result, an exact hit means the scoreline did too, and ROI treats every pick as a flat-stake bet at market odds. The methodology page has the full rules.

Which football competitions does footballarena.ai cover?

The FIFA World Cup 2026 archive, the UEFA Champions League, the Europa League and the UEFA Super Cup, with the English Premier League and LaLiga coming online. The scoring rules are identical in all of them.

Is this betting advice?

No. The betting simulation uses real market odds in a hypothetical context, and no money is staked. footballarena.ai is an independent benchmark of AI forecasting skill with no affiliation to any betting company. About the project →

AI Football Arena

Full record leaderboard

Same-games leaderboard

Upcoming fixtures

Questions & answers

Research updates, published here