Epoch
claude-sonnet-4-5-20250929
Made by Anthropic. Released 2025-09-29. Accessibility not recorded. Cheapest standard rate we hold: $15 per million output tokens, no context split.
What it costs
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| standard | text | $3 | $15 | $0.30 | licensed-index |
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 | 0.0±0.02 | 2026-08-06 | — |
| chess puzzles | Epoch AI | 0.1±0.03 | 2025-12-08 | 32K |
| ebr bench | Epoch AI | 0.0 | 2026-06-25 | — |
| frontiermath | Epoch AI | 0.2±0.02 | 2025-11-16 | 32K |
| frontiermath | Epoch AI | 0.1±0.02 | 2025-11-16 | — |
| frontiermath | Epoch AI | 0.1±0.02 | 2025-11-13 | 59K |
| frontiermath tier 4 | Epoch AI | 0.0±0.03 | 2025-10-22 | 32K |
| frontiermath tier 4 | Epoch AI | 0.0±0.02 | 2025-09-29 | — |
| frontiermath tier 4 v2 | Epoch AI | 0.0±0.02 | 2026-06-11 | 32K |
| frontiermath tiers 1 3 v2 | Epoch AI | 0.2±0.03 | 2026-06-11 | 32K |
| gpqa diamond | Epoch AI | 0.8±0.03 | 2025-10-28 | 59K |
| gpqa diamond | Epoch AI | 0.8±0.03 | 2025-10-28 | 16K |
| gpqa diamond | Epoch AI | 0.8±0.03 | 2025-10-21 | 32K |
| gpqa diamond | Epoch AI | 0.7±0.03 | 2025-09-29 | — |
| math level 5near its ceiling | Epoch AI | 1.0±0.00 | 2025-10-21 | 32K |
| mystery game puzzles | Epoch AI | 0.2±0.04 | 2026-07-25 | 48K |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.8±0.06 | 2025-10-28 | 59K |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.7±0.07 | 2025-10-28 | 16K |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.8±0.06 | 2025-10-21 | 32K |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.4±0.07 | 2025-09-29 | — |
| simpleqa verified | Epoch AI | 0.3±0.01 | 2026-08-27 | 59K |
| simpleqa verified | Epoch AI | 0.2±0.01 | 2026-08-10 | — |
| swe bench verified | Epoch AI | 0.7±0.02 | 2026-02-05 | — |
| ale bench | collected by Epoch AI | 796 | — | 32K |
| algotune | collected by Epoch AI | 1.5 | — | unknown |
| arc aginear its ceiling | collected by Epoch AI | 0.3 | — | — |
| arc aginear its ceiling | collected by Epoch AI | 0.3 | — | 1K |
| arc aginear its ceiling | collected by Epoch AI | 0.5 | — | 16K |
| arc aginear its ceiling | collected by Epoch AI | 0.5 | — | 8K |
| arc aginear its ceiling | collected by Epoch AI | 0.6 | — | 32K |
| arc agi 2 | collected by Epoch AI | 0.0 | — | — |
| arc agi 2 | collected by Epoch AI | 0.1 | — | 1K |
| arc agi 2 | collected by Epoch AI | 0.1 | — | 8K |
| arc agi 2 | collected by Epoch AI | 0.1 | — | 16K |
| arc agi 2 | collected by Epoch AI | 0.1 | — | 32K |
| critpt | collected by Epoch AI | 0.0 | — | unknown |
| cybench | collected by Epoch AI | 0.6 | — | — |
| cybench | collected by Epoch AI | 0.6 | — | unknown |
| deepresearchbench | collected by Epoch AI | 0.5 | — | 2K |
| deepresearchbench | collected by Epoch AI | 0.5 | — | low |
| dtbenchnear its ceiling | collected by Epoch AI | 0.8 | — | unknown |
| enigma eval | collected by Epoch AI | 0.1 | — | unknown |
| forecastbench | collected by Epoch AI | 61.9 | — | unknown |
| gdpval | collected by Epoch AI | 0.4 | — | — |
| gso | collected by Epoch AI | 0.1 | — | OpenHands |
| hle | collected by Epoch AI | 0.1 | — | unknown |
| lmca | collected by Epoch AI | 38.8 | — | unknown |
| metr time horizons | collected by Epoch AI | 122 | — | 16K |
| os world | collected by Epoch AI | 58.1 | — | claude-sonnet-4-5-20250929 (50 steps) |
| os world | collected by Epoch AI | 62.9 | — | claude-sonnet-4-5-20250929 (100 steps) |
| os world | collected by Epoch AI | 42.9 | — | claude-sonnet-4-5-20250929 (15 steps) |
| proofbench | collected by Epoch AI | 0.2 | — | unknown |
| rli | collected by Epoch AI | 0.0 | — | unknown |
| scicode | collected by Epoch AI | 0.4 | — | unknown |
| simplebench | collected by Epoch AI | 0.5 | — | unknown |
| simplebench | collected by Epoch AI | 0.5 | — | 12K |
| terminalbench | collected by Epoch AI | 0.4 | — | Claude Code |
| terminalbench | collected by Epoch AI | 0.5 | — | CAMEL-AI |
| terminalbench | collected by Epoch AI | 0.4 | — | Goose |
| terminalbench | collected by Epoch AI | 0.4 | — | Terminus 2 |
| terminalbench | collected by Epoch AI | 0.4 | — | MAYA |
| terminalbench | collected by Epoch AI | 0.4 | — | MAYA-V2 |
| terminalbench | collected by Epoch AI | 0.4 | — | OpenHands |
| terminalbench | collected by Epoch AI | 0.4 | — | Mini-SWE-Agent |
| vending bench 2 | collected by Epoch AI | 3839 | — | unknown |
| vpct | collected by Epoch AI | 0.4 | — | — |
| vpct | collected by Epoch AI | 0.4 | — | 32K |
| webdev arena | collected by Epoch AI | 1386 | — | — |
| webdev arena | collected by Epoch AI | 1391 | — | 32K |
| weirdml | collected by Epoch AI | 0.5 | — | 16K |
| weirdml | collected by Epoch AI | 0.5 | — | — |
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
- deprecation — 2026-09-29 (Active)
Every rate we hold · Why there is no single best model · Everything from Epoch