Google DeepMind · 2026-05-19 · seismic
Gemini 3.5 Flash — Google's I/O 2026 Model Rivals Flagship Frontier Models on Coding and Agentic Tasks at 4x the Speed
Google opened its Gemini 3.5 series at I/O 2026 with the Flash model, live now in the Gemini app, Search, Antigravity, and the API. Google says it matches large flagship models on coding and agentic tasks while running 4x faster.
Google's new Flash-tier model targets agentic coding at flagship quality and 4x the throughput.
Key specs
| Terminal bench 2.1 | 76.2% |
|---|---|
| Gdpval aa | 1656 Elo |
| Mcp atlas | 83.6% |
| Char xiv reasoning | 84.2% |
| Output speed | 4x faster than frontier models |
What is it?
Gemini 3.5 Flash is the first model in Google's Gemini 3.5 family, unveiled at I/O 2026. It is the speed-optimized tier, meant for high-volume, latency-sensitive work, but Google positions this generation as closing the gap with much larger flagship models on reasoning, coding, and agentic tasks. A bigger Gemini 3.5 Pro is in testing and due next month.
How does it work?
Like prior Flash models it trades some raw scale for inference speed, but Google reports frontier-level scores: 76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA, 83.6% on MCP Atlas, and 84.2% on the CharXiv multimodal reasoning benchmark. Google says output runs 4x faster than other frontier models, which is the main lever for long-horizon agent loops where many sequential model calls add up.
Why does it matter?
Agentic workflows make many model calls in sequence, so per-token speed and cost compound. A Flash-tier model that holds flagship-level accuracy lets developers run agents and coding tools without paying flagship latency or price. It shipped across Google's whole surface (app, Search, Antigravity, API, and enterprise) on day one.
Who is it for?
developers building agents and coding tools
Try it
Use model gemini-3.5-flash in the Gemini API via Google AI Studio (ai.dev)