How well will Grok 4 do on Frontier Math?
25
10kṀ110k
2026
15.3 percent
expected
1.8%
0 - 5%
0.5%
5 - 10%
96%
10 - 20%
1%
20 - 30%
0.8%
30 - 50%
0.5%
50 - 100%

The highest score of any version of Grok 4 on the Epoch AI dashboard for the FrontierMath benchmark, within 1 week of the first appearance of Grok 4 on the dashboard.

( https://epoch.ai/data/ai-benchmarking-dashboard )

Get
Ṁ1,000
to start trading!
Sort by:

@Fay42 Grok 4 just appeared on the Epoch AI dashboard

what happened?

@bh It got 12-14%

@Bayesian thanks, figured I’d missed it but didn’t see anything on the epoch.ai dashboard. i wonder if they will evaluate the heavy/multiagent version.

@Bayesian Why are we sure Grok 4 Heavy won't count? Description implies it would

bought Ṁ750 NO
bought Ṁ1 YES

@Bayesian where can you see the score? The link in the description doesn't appear to talk about grok4

@SimoneRomeo huh they tweeted about it but ig it's not on the site

bought Ṁ10 YES

@Bayesian @Fay42 do I understand right that if the score is not on their website within one week, the market should resolve 0%?

@SimoneRomeo I would have expected that to be NA.

@TimothyJohnson5c16 correct, N/A

I mean unless the creator disagrees and has a reasonable alternative

bought Ṁ1 YES

@Bayesian should be 0, it's the one that makes most sense across the options. There's no explanation about a potential N/A resolution in the criteria

@SimoneRomeo soo true

feel free to fill my NO limit order on 0-5% then

@Bayesian I may fill it once the author clarifies

@SimoneRomeo If it makes the most sense wouldn’t it be the most likely outcome? You won’t be able to fill my order once the creator clarifies

@Bayesian yes, correct

@SimoneRomeo The one week timer starts when it goes on the dashboard, as I say in the description, so it hasn't started yet.

© Manifold Markets, Inc.TermsPrivacy