OpenAI

gpt-5.2-2025-12-11

Made by OpenAI. Released 2025-12-11. 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 AI0.0±0.022026-07-13none
chess puzzlesEpoch AI0.5±0.052025-12-15xhigh
chess puzzlesEpoch AI0.4±0.052025-12-11high
chess puzzlesEpoch AI0.4±0.052025-12-11medium
chess puzzlesEpoch AI0.2±0.042025-12-11low
ebr benchEpoch AI0.22026-06-26xhigh
frontiermathEpoch AI0.4±0.032025-12-13xhigh
frontiermathEpoch AI0.4±0.032025-12-11medium
frontiermathEpoch AI0.4±0.032025-12-11high
frontiermathEpoch AI0.3±0.032025-12-11low
frontiermath tier 4Epoch AI0.2±0.062025-12-14xhigh
frontiermath tier 4Epoch AI0.2±0.052025-12-11medium
frontiermath tier 4Epoch AI0.2±0.062025-12-11high
frontiermath tier 4Epoch AI0.1±0.042025-12-11low
frontiermath tier 4 v2Epoch AI0.3±0.072026-06-11xhigh
frontiermath tiers 1 3 v2Epoch AI0.7±0.032026-06-11xhigh
gpqa diamondEpoch AI0.7±0.032026-07-13none
gpqa diamondEpoch AI0.9±0.022025-12-13xhigh
gpqa diamondEpoch AI0.8±0.022025-12-11low
gpqa diamondEpoch AI0.9±0.022025-12-11medium
gpqa diamondEpoch AI0.9±0.022025-12-11high
mystery game puzzlesEpoch AI0.2±0.042026-08-28medium
mystery game puzzlesEpoch AI0.1±0.032026-08-27low
mystery game puzzlesEpoch AI0.1±0.032026-08-27none
mystery game puzzlesEpoch AI0.2±0.042026-08-06high
otis mock aime 2024 2025near its ceilingEpoch AI0.6±0.072026-07-13none
otis mock aime 2024 2025near its ceilingEpoch AI1.0±0.032025-12-13xhigh
otis mock aime 2024 2025near its ceilingEpoch AI1.0±0.032025-12-11high
otis mock aime 2024 2025near its ceilingEpoch AI0.9±0.032025-12-11medium
otis mock aime 2024 2025near its ceilingEpoch AI0.8±0.052025-12-11low
simpleqa verifiedEpoch AI0.3±0.012026-08-27medium
simpleqa verifiedEpoch AI0.3±0.012026-08-27low
simpleqa verifiedEpoch AI0.4±0.022026-08-27xhigh
simpleqa verifiedEpoch AI0.3±0.022026-08-27high
swe bench verifiedEpoch AI0.7±0.022026-02-12high
ale benchcollected by Epoch AI1250medium
ale benchcollected by Epoch AI1294high
algotunecollected by Epoch AI2.0medium
apex agentscollected by Epoch AI0.5xhigh
apex agentscollected by Epoch AI0.4unknown
apex agentscollected by Epoch AI0.4high
arc aginear its ceilingcollected by Epoch AI0.7medium
arc aginear its ceilingcollected by Epoch AI0.1unknown
arc aginear its ceilingcollected by Epoch AI0.1none
arc aginear its ceilingcollected by Epoch AI0.6low
arc aginear its ceilingcollected by Epoch AI0.8high
arc aginear its ceilingcollected by Epoch AI0.9xhigh
arc agi 2collected by Epoch AI0.1low
arc agi 2collected by Epoch AI0.5xhigh
arc agi 2collected by Epoch AI0.0unknown
arc agi 2collected by Epoch AI0.4high
arc agi 2collected by Epoch AI0.3medium
cl benchcollected by Epoch AI0.2high
cl benchcollected by Epoch AI0.2unknown
deepresearchbenchcollected by Epoch AI0.4low
dtbenchnear its ceilingcollected by Epoch AI0.9xhigh
enigma evalcollected by Epoch AI0.1unknown
forecastbenchcollected by Epoch AI60.1unknown
gdpvalcollected by Epoch AI0.5none
gsocollected by Epoch AI0.3OpenHands
hlecollected by Epoch AI0.3unknown
lmcacollected by Epoch AI43.9xhigh
metr time horizonscollected by Epoch AI352high
proofbenchcollected by Epoch AI0.1xhigh
rlicollected by Epoch AI0.0medium
rlicollected by Epoch AI0.0unknown
simplebenchcollected by Epoch AI0.5unknown
simplebenchcollected by Epoch AI0.5high
terminalbenchcollected by Epoch AI0.6Mux
terminalbenchcollected by Epoch AI0.6Codex CLI
terminalbenchcollected by Epoch AI0.6Droid
terminalbenchcollected by Epoch AI0.5Terminus 2
vending bench 2collected by Epoch AI3591unknown
vpctcollected by Epoch AI0.7high
vpctcollected by Epoch AI0.8xhigh
webdev arenacollected by Epoch AI1480high
webdev arenacollected by Epoch AI1418unknown
webdev arenacollected by Epoch AI1394medium
weirdmlcollected by Epoch AI0.5low
weirdmlcollected by Epoch AI0.6medium
weirdmlcollected by Epoch AI0.5none
weirdmlcollected by Epoch AI0.7xhigh

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.

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