Which Benchmarks will OpenAI show results from GPT-5 on, when it is announced?
12
1kṀ2875
2026
94%
SimpleQA
11%
GSM8K
71%
HumanEval
82%
MMLU
83%
GPQA
62%
MATH
41%
MGSM
29%
DROP
47%
Big-Bench-Hard
87%
SWE-Bench

Some flexibility on variations of specific benchmarks. eg SWE-Bench-Hard would resolve SWE-Bench YES.

  • Update 2025-05-11 (PST) (AI summary of creator comment): The benchmarks must be those that GPT-5 is benchmarked against by OpenAI.

Get
Ṁ1,000
to start trading!
© Manifold Markets, Inc.TermsPrivacy