Groq
qwen3.8-27b
Made by Alibaba. Released 2026-08-14. Accessibility not recorded. Cheapest standard rate we hold: $4 per million output tokens, no context split.
What it costs
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| standard | text | $0.80 | $4 | — | registry |
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 |
|---|---|---|---|---|
| critpt | collected by Epoch AI | 0.1 | — | xhigh |
| critpt | collected by Epoch AI | 0.0 | — | medium |
| critpt | collected by Epoch AI | 0.0 | — | none |
| critpt | collected by Epoch AI | 0.0 | — | low |
| proofbench | collected by Epoch AI | 0.2 | — | — |
| scicode | collected by Epoch AI | 0.4 | — | low |
| scicode | collected by Epoch AI | 0.4 | — | none |
| scicode | collected by Epoch AI | 0.4 | — | medium |
| scicode | collected by Epoch AI | 0.4 | — | xhigh |
| surface evolver bench | collected by Epoch AI | 0.5 | — | xhigh |
| webdev arena | collected by Epoch AI | 1595 | — | — |
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 · Everything from Groq