Gemini 3.5 Flash
Gemini 3.5 Flash, launched at I/O 2026, beats 3.1 Pro on agentic benchmarks, with 76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA and 83.6% on MCP Atlas.
Kicks off the 3.5 family, reportedly 4x faster in output tokens per second than other frontier models, and became the default in the Gemini app and AI Mode. Pairs with the Antigravity harness for parallel subagents. A 3.5 Pro was promised "next month" but had not shipped by 2026-10-04.
- Date
- Tuesday, 19 May 2026
- Lab
- Google DeepMind
- Kind
- model
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| Terminal-Bench 2.1 | 76.2% company-reported | company |
| GDPval-AA | 1656 Elo company-reported | company |
| MCP Atlas | 83.6% company-reported | company |
| CharXiv Reasoning | 84.2% company-reported | company |
Speed claim is Google's and is based on output tokens per second. GA and developer release 2026-05-19; computer use added 2026-06-24. Successors were 3.6 Flash (2026-07-21), 3.7 (2026-08-13) and 3.8 (2026-09-02).
Sources
- deepmind.google/blog/gemini-3-5-frontier-intelligence-with-action/
- ai.google.dev/gemini-api/docs/changelog
- deepmind.google/models/gemini/pro/
This record was checked against its sources on 6 October 2026. How we check