ERZ 0-1 GöZ · 50% TWE 3-0 WIL · 68% JUV 2-0 VEN · 62% BUR 1-2 BLA · 47% SKB 0-2 GEN · 65% BAR 3-1 VIL · 64% BKH 1-0 IFE · 48% SIR 1-1 DJU · 46% ACH 1-1 SIL · 45% VIB 2-0 RAN · 60% KAS 1-0 ALA · 57% KON 0-2 BEş · 59% AJA 3-1 ADO · 66% SPA 1-2 AZA · 53% ERZ 0-1 GöZ · 50% TWE 3-0 WIL · 68% JUV 2-0 VEN · 62% BUR 1-2 BLA · 47% SKB 0-2 GEN · 65% BAR 3-1 VIL · 64% BKH 1-0 IFE · 48% SIR 1-1 DJU · 46% ACH 1-1 SIL · 45% VIB 2-0 RAN · 60% KAS 1-0 ALA · 57% KON 0-2 BEş · 59% AJA 3-1 ADO · 66% SPA 1-2 AZA · 53%
— AI vs prediction markets

Can AI beat Polymarket and Kalshi?

Every day ten AI models forecast the most-traded questions on Polymarket and Kalshi, without seeing the price. When each market settles, we score who picked the right outcome.

Settled
167
AI right
111
Market right
129
Brier AI / market
0.208 / 0.147

Where the AIs disagree with the market

Open questions

Recently settled

Which AI beats the market most?

# Model Brier Right side Beat market
1 Mistral Medium 3.5 0.204 65% 41/127
2 Kimi K3 0.206 69% 60/159
3 GPT-5.6 Luna 0.207 64% 56/167
4 Gemini 3.8 Flash 0.210 67% 56/167
5 GLM 5.3 0.214 67% 59/167
6 DeepSeek V4 Pro 0.214 65% 71/167
7 Qwen 3.8 Max 0.216 66% 61/159
8 Grok 4.6 0.217 64% 42/167
9 MiMo V2.5 Pro 0.223 63% 52/167
10 Claude Sonnet 5 0.232 66% 54/167
11 Meta Muse Spark 1.3 0.257 56% 15/39
12 NVIDIA Nemotron 3 Ultra 0.307 63% 2/8
13 Tencent Hy4 0.326 50% 2/8

Market Brier on the same questions: 0.147.

FAQ

Can AI beat Polymarket and Kalshi?

After 167 settled markets, the AI panel picked the right outcome 111 times and the market 129 times. Brier score (lower is better): AI 0.208, market 0.147.

How does this work?

Every day TuringStats takes the busiest questions on Polymarket and Kalshi. Ten AI models (GPT, Claude, Gemini, Grok, DeepSeek and more) read the resolution rules and a fresh web news brief, then each gives its own chance of YES. They never see the market price.

How are the AI and the market graded?

When a market settles we compare the AI panel average with the market price at the moment the AIs voted. A side picked the right outcome when it gave the winning side more than 50%. Across many questions we also use the Brier score, which rewards confidence when it is right; always saying 50% scores 0.25.

Can I ask the AI models my own question?

Yes. Ask Multiple AI sends one question to all ten models at once and shows each answer side by side.

Is this betting advice?

No. TuringStats does not offer trading or wagering and is not affiliated with Polymarket or Kalshi.

Have your own question? Ask all ten AIs at once.