OpenAI
gpt-3.5-turbo-1106
Made by OpenAI. Released 2023-11-06. Accessibility not recorded. Cheapest standard rate we hold: $2 per million output tokens, at long context.
What it costs
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| batch | text · long | $1 | $2 | — | seller |
| batch | text · short | $1 | $2 | — | seller |
| standard | text · long | $1 | $2 | — | seller |
| standard | text · short | $1 | $2 | — | 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 |
|---|---|---|---|---|
| gpqa diamond | Epoch AI | 0.3±0.02 | 2025-01-27 | — |
| math level 5near its ceiling | Epoch AI | 0.2±0.01 | 2025-01-27 | — |
| adversarial nli | collected by Epoch AI | 0.6 | — | — |
| arc ai2 | collected by Epoch AI | 0.9 | — | — |
| mmlu | collected by Epoch AI | 0.7 | — | — |
| open book qa | collected by Epoch AI | 0.9 | — | — |
| trivia qa | collected by Epoch AI | 0.9 | — | — |
| wino grande | collected by Epoch AI | 0.7 | — | — |
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 — 2026-09-28, replaced by gpt-5.6-terra
Every rate we hold · Why there is no single best model · Everything from OpenAI