OpenAI

gpt-5-2025-08-07

Made by OpenAI. Released 2025-08-07. Accessibility not recorded. We hold no price for this model: no seller we can read lists it any more, so what follows is what was measured, not what it cost.

What it scores, by house

BenchmarkHouseScoreEvaluatedScaffold
chess puzzlesEpoch AI0.2±0.042026-07-15low
chess puzzlesEpoch AI0.3±0.052026-07-15medium
chess puzzlesEpoch AI0.2±0.042026-07-15minimal
chess puzzlesEpoch AI0.4±0.052025-12-08high
ebr benchEpoch AI0.12026-06-29high
frontiermathEpoch AI0.3±0.032025-11-13high
frontiermathEpoch AI0.3±0.032025-11-13medium
frontiermath tier 4Epoch AI0.1±0.052025-10-30high
frontiermath tier 4Epoch AI0.1±0.042025-08-07medium
frontiermath tier 4 v2Epoch AI0.2±0.072026-06-11high
frontiermath tiers 1 3 v2Epoch AI0.4±0.032026-08-27low
frontiermath tiers 1 3 v2Epoch AI0.2±0.022026-08-27minimal
frontiermath tiers 1 3 v2Epoch AI0.6±0.032026-06-10high
gpqa diamondEpoch AI0.7±0.032026-07-20minimal
gpqa diamondEpoch AI0.9±0.022025-10-29high
gpqa diamondEpoch AI0.9±0.022025-08-07medium
math level 5near its ceilingEpoch AI1.0±0.002025-10-29high
math level 5near its ceilingEpoch AI1.0±0.002025-08-20medium
mystery game puzzlesEpoch AI0.1±0.032026-08-27minimal
mystery game puzzlesEpoch AI0.2±0.042026-08-27low
mystery game puzzlesEpoch AI0.1±0.042026-08-27medium
mystery game puzzlesEpoch AI0.2±0.042026-08-05high
otis mock aime 2024 2025near its ceilingEpoch AI0.5±0.082026-07-20minimal
otis mock aime 2024 2025near its ceilingEpoch AI0.9±0.042025-10-29high
otis mock aime 2024 2025near its ceilingEpoch AI0.9±0.042025-08-07medium
simpleqa verifiedEpoch AI0.5±0.022026-08-27high
swe bench verifiedEpoch AI0.7±0.022026-02-06high
swe bench verifiedEpoch AI0.7±0.022026-02-05medium
aider polyglotcollected by Epoch AI88.0high
aider polyglotcollected by Epoch AI86.7medium
aider polyglotcollected by Epoch AI81.3low
ale benchcollected by Epoch AI808minimal
ale benchcollected by Epoch AI1162high
algotunecollected by Epoch AI1.7high
apex agentscollected by Epoch AI0.3high
apex agentscollected by Epoch AI0.3unknown
arc aginear its ceilingcollected by Epoch AI0.6medium
arc aginear its ceilingcollected by Epoch AI0.1minimal
arc aginear its ceilingcollected by Epoch AI0.7high
arc aginear its ceilingcollected by Epoch AI0.4low
arc agi 2collected by Epoch AI0.1high
arc agi 2collected by Epoch AI0.0minimal
arc agi 2collected by Epoch AI0.0low
arc agi 2collected by Epoch AI0.1medium
balrogcollected by Epoch AI0.3minimal
critptcollected by Epoch AI0.0minimal
critptcollected by Epoch AI0.1high
deepresearchbenchcollected by Epoch AI0.5low
deepresearchbenchcollected by Epoch AI0.5high
deepresearchbenchcollected by Epoch AI0.5medium
deepresearchbenchcollected by Epoch AI0.5minimal
dtbenchnear its ceilingcollected by Epoch AI0.9high
enigma evalcollected by Epoch AI0.1unknown
fictionlivebenchcollected by Epoch AI1.0medium
forecastbenchcollected by Epoch AI61.4unknown
gdpvalcollected by Epoch AI0.3medium
geobenchcollected by Epoch AI3937medium
gsocollected by Epoch AI0.1OpenHands
hlecollected by Epoch AI0.3high
lech mazur writingcollected by Epoch AI8.6medium
lmcacollected by Epoch AI40.0high
metr time horizonscollected by Epoch AI203high
metr time horizonscollected by Epoch AI137medium
proofbenchcollected by Epoch AI0.2high
rlicollected by Epoch AI0.0unknown
scicodecollected by Epoch AI0.4unknown
simplebenchcollected by Epoch AI0.6high
terminalbenchcollected by Epoch AI0.3Mini-SWE-Agent
terminalbenchcollected by Epoch AI0.5Codex CLI
terminalbenchcollected by Epoch AI0.4OpenHands
terminalbenchcollected by Epoch AI0.4Terminus 2
vpctcollected by Epoch AI0.6medium
vpctcollected by Epoch AI0.7high
webdev arenacollected by Epoch AI1419medium
weirdmlcollected by Epoch AI0.6high

These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.

What retires, and when

  • deprecation2026-12-11, replaced by gpt-5.6-sol

Every rate we hold · Why there is no single best model ·