Frontier Log

OpenAIModel

GPT-5

Made by
OpenAI
Released
Announced in
Introducing GPT-5openai.com
Event
GPT-5
API record
/api/v1/models/gpt-5

Benchmark results

BenchmarkScoreSource check
Humanity's Last ExamNo tools · with thinking; full set; pass@124.8%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 2 other lab documents: 1 2
without thinking; full set; pass@16.3%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Humanity's Last ExamWith tools · with thinking; Python and search with a blocklist; full set; pass@135.2%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
SWE-bench Verified477 of the 500 tasks74.9%
CheckedA person checked this number against its source on 16 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 4 other lab documents, the first 3 linked: 1 2 3
high reasoning effort; all 500 problems · reported later, in Introducing GPT-5.1 for developers72.8%
CorroboratedThe same number is printed in 2 other lab documents: 1 2
GPQA Diamondwith thinking; no tools; pass@185.7%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
without thinking; no tools; pass@177.8%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
with thinking; Python; pass@187.3%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
BrowseCompwith thinking; pass@154.9%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
AIME 2025No tools · with thinking; pass@194.6%
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 4 other lab documents, the first 3 linked: 1 2 3
without thinking; pass@161.9%
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
AIME 2025With tools · with thinking; Python; pass@1; OpenAI says not to compare with-tools scores to models without tools99.6%
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 2 other lab documents: 1 2
without thinking; Python; pass@171.0%
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Sources

Reported by OpenAI

Events

Stories

All models