Release2026-09-02

Google DeepMind published the model card for Gemini 3.8 Flash. The model builds on Gemini 3.7 Flash and reuses its architecture and training data. Google reports gains in software engineering and agent-style knowledge work. Users can still set effort levels to trade quality against cost and speed. It accepts text, images, audio and video, and returns text only. Distribution covers the Gemini app, AI Studio, the Gemini API, Gemini Enterprise Agent Platform, AI Mode and Antigravity. Google lists hallucinations, occasional slowness and heavy token use at high effort as limits. Safety scores match 3.7 Flash, though non-English safety regressed slightly. Google says the model reaches no tracked capability levels under its Frontier Safety Framework.

What changed

Gemini 3.7 Flash was the current Flash model in the Gemini 3 family.

What it unlocks

Running lower-cost production agents on software engineering and knowledge tasks with a chosen effort level.

  • 1M token context window
  • 64K token output
  • knowledge cutoff March 2026
  • multilingual safety +5.4pp

What you need to act on it

  • access to the Gemini app, AI Studio, Gemini API, Gemini Enterprise Agent Platform, AI Mode or Antigravity

Send this to someone who needs it

Shares the story and its sources. Nothing about you.

What does this mean for your job?

This is the story as everyone gets it. Once a week we send you the version written for your role — what changed, why it matters for the work you actually do, and one thing to try. Free while we tune it.