Gemini 3.5 Flash-Lite
Made by Google DeepMind. Released 2026-07-21. Accessibility not recorded. Cheapest standard rate we hold: $2.50 per million output tokens, no context split.
What it costs
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| batch | text | $0.15 | $1.25 | — | seller |
| flex | text | $0.15 | $1.25 | — | seller |
| priority | text | $0.54 | $4.50 | — | seller |
| standard | text | $0.30 | $2.50 | — | 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 | 0.2±0.04 | 2026-08-06 | high |
| chess puzzles | Epoch AI | 0.2±0.04 | 2026-08-06 | minimal |
| chess puzzles | Epoch AI | 0.2±0.04 | 2026-08-06 | low |
| frontiermath tier 4 v2 | Epoch AI | 0.0±0.00 | 2026-08-02 | high |
| frontiermath tiers 1 3 v2 | Epoch AI | 0.3±0.03 | 2026-08-02 | high |
| gpqa diamond | Epoch AI | 0.7±0.03 | 2026-08-06 | minimal |
| gpqa diamond | Epoch AI | 0.8±0.03 | 2026-08-06 | low |
| gpqa diamond | Epoch AI | 0.8±0.03 | 2026-08-06 | high |
| mystery game puzzles | Epoch AI | 0.1±0.03 | 2026-08-05 | minimal |
| mystery game puzzles | Epoch AI | 0.2±0.04 | 2026-08-05 | low |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.5±0.08 | 2026-08-06 | minimal |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.7±0.07 | 2026-08-06 | high |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.6±0.07 | 2026-08-06 | low |
| ale bench | collected by Epoch AI | 765 | — | high |
| arc aginear its ceiling | collected by Epoch AI | 0.2 | — | low |
| arc aginear its ceiling | collected by Epoch AI | 0.5 | — | high |
| arc aginear its ceiling | collected by Epoch AI | 0.3 | — | medium |
| arc aginear its ceiling | collected by Epoch AI | 0.1 | — | minimal |
| arc agi 2 | collected by Epoch AI | 0.0 | — | minimal |
| arc agi 2 | collected by Epoch AI | 0.0 | — | low |
| arc agi 2 | collected by Epoch AI | 0.1 | — | medium |
| arc agi 2 | collected by Epoch AI | 0.1 | — | high |
| critpt | collected by Epoch AI | 0.0 | — | — |
| dtbenchnear its ceiling | collected by Epoch AI | 0.8 | — | high |
| gdp pdf | collected by Epoch AI | 0.1 | — | high |
| lmca | collected by Epoch AI | 37.6 | — | high |
| proofbench | collected by Epoch AI | 0.1 | — | — |
| scicode | collected by Epoch AI | 0.4 | — | — |
| webdev arena | collected by Epoch AI | 1449 | — | — |
| weirdml | collected by Epoch AI | 0.4 | — | 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.
Every rate we hold · Why there is no single best model · Everything from Google