Prediction Arena

Can AI predict the future? 6 models. $60K. Running in a loop until $0 in the bank. The players to start? Only the best from @OpenAI, @AnthropicAI, @xai, @Zai_org, and @GoogleDeepMind Find out now at predictionarena.ai

Fun fact, GPT 5.5 is very good at Game Dev Game Dev is the notable category where @OpenAI consistently beats out @AnthropicAI's Claude models Upon code inspection, our @Designarena team found that GPT 5.5's frontend verbosity plays in its favor for game dev - it consistently created games with the most functional features Congrats to @OpenAI for establishing the new Game Dev frontier!

@SDBarrington @KamrynOhly @AnthropicAI @Polymarket Both, paper here: arxiv.org/abs/2604.07355

@KamrynOhly @AnthropicAI @Polymarket @predictionbench Paper trading or real?

Our team is stunned. We gave Claude Opus 4.6 by @AnthropicAI $10k to trade on @Polymarket. It’s now has an account value of $70,614.59. This is a new era of model performance in trading and predicting outcomes in the face of uncertainty. @predictionbench

@Mayhem4Markets @KamrynOhly @AnthropicAI @Polymarket 5 of the 9 models use real cash

@KamrynOhly @AnthropicAI @Polymarket @predictionbench Hold the phone, you let Grok have real money and the rest of the models "paper traded"? 💀

@mpapas @KamrynOhly @AnthropicAI @Polymarket Not quite :)

@KamrynOhly @AnthropicAI @Polymarket @predictionbench $82,743.23 in model usage costs?

@stoked_on_waves @KamrynOhly @AnthropicAI @Polymarket Thanks for flagging - the link we shared pointed to the wrong market. The actual “Israel x Lebanon Diplomatic Meeting” market recently closed, so it no longer appears in Polymarket search, though it still exists in their Gamma API under the original slug.

@predictionbench @KamrynOhly @AnthropicAI @Polymarket the link you provided is for a diplomatic meeting between israel and hezbollah by april 26. This market trades at 1% with 15k volume. the chart of the performance shows a position value of ~50k and a price >99 cents

@stoked_on_waves @KamrynOhly @AnthropicAI @Polymarket We're pushing a fix for this bug, here's the url for reference polymarket.com/event/israel-x…

@KamrynOhly @AnthropicAI @Polymarket @predictionbench and the link to the market for its supposed to have 77% of its money in doesn't work, its not clear if that market, exists

@DeltaFlat @KamrynOhly @AnthropicAI @Polymarket

@KamrynOhly @AnthropicAI @Polymarket @predictionbench x.com/deltaflat/stat…

@abcampbell @NikLentz @KamrynOhly @AnthropicAI @Polymarket No problem!

@McNolf @KamrynOhly @AnthropicAI @Polymarket Here's the full methodology: arxiv.org/html/2604.0735…

@McNolf @KamrynOhly @AnthropicAI @Polymarket Haha for the live trading, yes! But devastatingly not for the paper trading:

For folks asking about the active positions...

@gsmmtt @darshil @KamrynOhly @AnthropicAI @Polymarket 🫡

@darshil @KamrynOhly @AnthropicAI @Polymarket @predictionbench I skimmed the paper, seems to be because on kalshi the models are forced into tougher markets, but in polymarket they can specialize in categories or questions where they seem to have edge because of better discovery

@NikLentz @abcampbell @KamrynOhly @AnthropicAI @Polymarket Yes! And more on our website

@abcampbell @KamrynOhly @AnthropicAI @Polymarket @predictionbench Def would be an important detail to have.

@danielkempe @KamrynOhly @AnthropicAI @Polymarket It's hard to say because the time horizons are different, so absolute returns might not be the best metric here We're doing more analysis that tries to control for these variances and will follow up soon! Please send us more questions to dig into

@predictionbench @KamrynOhly @AnthropicAI @Polymarket And is there a major difference in returns?

@McNolf @KamrynOhly @AnthropicAI @Polymarket We already did lol

@KamrynOhly @AnthropicAI @Polymarket @predictionbench ok put $50k of your money into your fund

@imdecentralized @KamrynOhly @AnthropicAI @Polymarket We have both real and paper trading (as you can imagine, for monetary sustainability). The system currently runs with $50,000 USD in real capital, and we're considering expanding. Find our full methodology for live trading here: arxiv.org/html/2604.0735…

@KamrynOhly @AnthropicAI @Polymarket @predictionbench Paper trading in illiquid markets is not a benchmark

@atajik @KamrynOhly More on this soon!

@KamrynOhly @predictionbench curious why Opus is at $7k on Kalshi but at $70k on Polymarket?

@skill @KamrynOhly @AnthropicAI @Polymarket We have both real and paper trading (as you can imagine, for monetary sustainability). The system currently runs with $50,000 USD in real capital, and we're considering expanding. Find our full methodology for live trading here: arxiv.org/html/2604.0735…

@KamrynOhly @AnthropicAI @Polymarket @predictionbench No real money, skill issue.

@Jake_Joseph @KamrynOhly @AnthropicAI @Polymarket Thank you!

@KamrynOhly @AnthropicAI @Polymarket @predictionbench This is incredibly interesting. Hope you continue to share more.

@flip_coinz @KamrynOhly @AnthropicAI @Polymarket Already done! arxiv.org/html/2604.0735…

@KamrynOhly @AnthropicAI @Polymarket @predictionbench paper trading ain’t real. put real money up or shut up