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
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| batch | text · long | $10 | $38 | $1 | seller |
| batch | text · short | $5 | $25 | $0.50 | seller |
| fast | text · long | $40 | $150 | $4 | seller |
| fast | text · short | $20 | $100 | $2 | seller |
| flex | text · long | $10 | $38 | $1 | seller |
| flex | text · short | $5 | $25 | $0.50 | seller |
| standard | text · long | $20 | $75 | $2 | seller |
| standard | text · short | $10 | $50 | $1 | seller |
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
| Benchmark | House | Score | Evaluated | Scaffold |
|---|---|---|---|---|
| chess puzzles | Epoch AI | 0.7±0.05 | 2026-08-30 | max |
| ebr bench | Epoch AI | 0.8±0.08 | 2026-08-30 | max |
| frontiermath erdos | Epoch AI | 0.0±0.02 | 2026-08-28 | max |
| frontiermath tier 4 v2 | Epoch AI | 1.0±0.02 | 2026-08-30 | high |
| frontiermath tier 4 v2 | Epoch AI | 0.8±0.06 | 2026-08-30 | none |
| frontiermath tier 4 v2 | Epoch AI | 0.9±0.05 | 2026-08-30 | low |
| frontiermath tier 4 v2 | Epoch AI | 1.0±0.02 | 2026-08-30 | medium |
| frontiermath tier 4 v2 | Epoch AI | 1.0±0.02 | 2026-08-30 | max |
| frontiermath tier 4 v2 | Epoch AI | 1.0±0.02 | 2026-08-30 | xhigh |
| frontiermath tiers 1 3 v2 | Epoch AI | 0.9±0.01 | 2026-08-30 | max |
| gpqa diamond | Epoch AI | 1.0±0.01 | 2026-08-30 | max |
| mirrorcode | Epoch AI | 0.5 | 2026-08-30 | high |
| mystery game puzzles | Epoch AI | 0.8±0.04 | 2026-08-30 | max |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 1.0±0.00 | 2026-08-30 | max |
| simpleqa verified | Epoch AI | 0.8±0.01 | 2026-08-30 | max |
| apex agents | collected by Epoch AI | 0.6 | — | unknown |
| arc aginear its ceiling | collected by Epoch AI | 1.0 | — | low |
| arc aginear its ceiling | collected by Epoch AI | 1.0 | — | medium |
| arc aginear its ceiling | collected by Epoch AI | 1.0 | — | max |
| arc aginear its ceiling | collected by Epoch AI | 1.0 | — | xhigh |
| arc aginear its ceiling | collected by Epoch AI | 1.0 | — | high |
| arc aginear its ceiling | collected by Epoch AI | 0.9 | — | none |
| arc agi 2 | collected by Epoch AI | 0.9 | — | high |
| arc agi 2 | collected by Epoch AI | 0.9 | — | max |
| arc agi 2 | collected by Epoch AI | 0.9 | — | xhigh |
| arc agi 2 | collected by Epoch AI | 0.9 | — | medium |
| arc agi 2 | collected by Epoch AI | 0.6 | — | none |
| arc agi 2 | collected by Epoch AI | 0.9 | — | low |
| critpt | collected by Epoch AI | 0.2 | — | none |
| critpt | collected by Epoch AI | 0.3 | — | xhigh |
| critpt | collected by Epoch AI | 0.3 | — | medium |
| critpt | collected by Epoch AI | 0.3 | — | low |
| critpt | collected by Epoch AI | 0.3 | — | max |
| critpt | collected by Epoch AI | 0.3 | — | high |
| deepswe | collected by Epoch AI | 0.7 | — | mini-swe-agent |
| frontiercode | collected by Epoch AI | 0.5 | — | codex |
| proofbench | collected by Epoch AI | 1.0 | — | unknown |
| scicode | collected by Epoch AI | 0.5 | — | medium |
| scicode | collected by Epoch AI | 0.5 | — | none |
| scicode | collected by Epoch AI | 0.5 | — | low |
| scicode | collected by Epoch AI | 0.6 | — | high |
| scicode | collected by Epoch AI | 0.6 | — | xhigh |
| scicode | collected by Epoch AI | 0.6 | — | max |
| webdev arena | collected by Epoch AI | 1797 | — | max |
| weirdml | collected by Epoch AI | 0.9 | — | high |
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