Gemini 3.7 Flash
by Google / Google DeepMind
August 2026 Flash release — 50% cheaper tokens, a free API tier, and coding benchmarks above Claude and GPT.
Part of the Gemini 3 series.
Timeline
| Date | Event |
|---|---|
| 2026-08-08 | First reports |
| 2026-08-13 | 50% cheaper tokens; reported to beat Claude and GPT on coding |
| 2026-08-17 | Broad testing — “Google’s best model yet” |
| 2026-08-23 | Fully free API tier plus a free Antigravity tier |
| 2026-08-26 | Framed as changing agent economics |
Notes
The most consequential Gemini release of 2026, and the reason is price rather than capability: a 50% token cut plus a genuinely free API tier changed what agent loops cost to run.
Several sources noted the free tier was under-discussed relative to its impact.
This is the third Flash iteration inside roughly fourteen weeks (3.5 → 3.6 → 3.7) — an unusually
rapid cadence even by 2026 standards.