OpenAI

gpt-6-astra

Made by OpenAI. Released 2026-09-03. Accessibility not recorded. Cheapest standard rate we hold: $50 per million output tokens, at short context.

What it costs

TierKindInOutCached inFrom
batchtext · long$10$38$1seller
batchtext · short$5$25$0.50seller
fasttext · long$40$150$4seller
fasttext · short$20$100$2seller
flextext · long$10$38$1seller
flextext · short$5$25$0.50seller
standardtext · long$20$75$2seller
standardtext · short$10$50$1seller

per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.

What it scores, by house

BenchmarkHouseScoreEvaluatedScaffold
chess puzzlesEpoch AI0.7±0.052026-08-30max
ebr benchEpoch AI0.8±0.082026-08-30max
frontiermath erdosEpoch AI0.0±0.022026-08-28max
frontiermath tier 4 v2Epoch AI1.0±0.022026-08-30high
frontiermath tier 4 v2Epoch AI0.8±0.062026-08-30none
frontiermath tier 4 v2Epoch AI0.9±0.052026-08-30low
frontiermath tier 4 v2Epoch AI1.0±0.022026-08-30medium
frontiermath tier 4 v2Epoch AI1.0±0.022026-08-30max
frontiermath tier 4 v2Epoch AI1.0±0.022026-08-30xhigh
frontiermath tiers 1 3 v2Epoch AI0.9±0.012026-08-30max
gpqa diamondEpoch AI1.0±0.012026-08-30max
mirrorcodeEpoch AI0.52026-08-30high
mystery game puzzlesEpoch AI0.8±0.042026-08-30max
otis mock aime 2024 2025near its ceilingEpoch AI1.0±0.002026-08-30max
simpleqa verifiedEpoch AI0.8±0.012026-08-30max
apex agentscollected by Epoch AI0.6unknown
arc aginear its ceilingcollected by Epoch AI1.0low
arc aginear its ceilingcollected by Epoch AI1.0medium
arc aginear its ceilingcollected by Epoch AI1.0max
arc aginear its ceilingcollected by Epoch AI1.0xhigh
arc aginear its ceilingcollected by Epoch AI1.0high
arc aginear its ceilingcollected by Epoch AI0.9none
arc agi 2collected by Epoch AI0.9high
arc agi 2collected by Epoch AI0.9max
arc agi 2collected by Epoch AI0.9xhigh
arc agi 2collected by Epoch AI0.9medium
arc agi 2collected by Epoch AI0.6none
arc agi 2collected by Epoch AI0.9low
critptcollected by Epoch AI0.2none
critptcollected by Epoch AI0.3xhigh
critptcollected by Epoch AI0.3medium
critptcollected by Epoch AI0.3low
critptcollected by Epoch AI0.3max
critptcollected by Epoch AI0.3high
deepswecollected by Epoch AI0.7mini-swe-agent
frontiercodecollected by Epoch AI0.5codex
proofbenchcollected by Epoch AI1.0unknown
scicodecollected by Epoch AI0.5medium
scicodecollected by Epoch AI0.5none
scicodecollected by Epoch AI0.5low
scicodecollected by Epoch AI0.6high
scicodecollected by Epoch AI0.6xhigh
scicodecollected by Epoch AI0.6max
webdev arenacollected by Epoch AI1797max
weirdmlcollected by Epoch AI0.9high

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.

Headlines that name it

Matched by the model's name appearing in our headline, literally. There is no link from a story to a model in our data, so this is a search result rather than a claim that the story is about it.

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