Gemini 3.7 Flash

ASSESSMENT

Gemini 3.7 Flash is Google's newest fast multimodal tier, announced August 13, 2026 — the second Flash refresh in under a month, after 3.6 Flash on July 21. It keeps the Gemini line's full native multimodality and 1M-token context, and posts sizable coding and agentic gains over 3.6: Google reports DeepSWE v1.1 rising from 49.0% to 65.3%, FrontierCode 1.1 Main from 34.4% to 43.6%, and a Code Arena web-dev Elo of 1588.

Those figures are vendor-reported. The one independent read so far, from Artificial Analysis, places 3.7 Flash about four points above 3.6 Flash and on its intelligence-versus-speed Pareto frontier — but a full independent index is not yet settled. Pricing is introductory: $0.75 / $3.75 per million tokens through the end of 2026, scheduled to double to $1.50 / $7.50 on January 1, 2027, so model longer-term cost at the higher rate. Weights are closed.

For latency-sensitive, high-volume, or multimodal work that also leans on coding and agentic reliability, 3.7 Flash is Google's current Flash pick and a clear step up from 3.5 Flash. For the hardest pure reasoning the independently ranked flagships still have the stronger claim, and for Google's top reasoning tier the model to build on remains Gemini 3.1 Pro.