Three weeks. That's the gap between 3.6 Flash and this release, and the pitch is aimed squarely at people wiring models into agent loops and coding tools. The headline is the coding jump — the company reports FrontierCode 1.1 climbing to 43.6% from 34.4% and DeepSWE v1.1 to 65.3% from 49.0% — alongside the familiar 1M-token input window and up to 64K tokens of output per response. The part that'll actually move your bill is the introductory price: $0.75 per million input tokens and $3.75 output, running through the end of the year, roughly half what 3.6 cost. It's live in the Gemini API via AI Studio, in Android Studio, and through the enterprise agent platform. If you're already benchmarking Flash-tier models for an agent, this one earns a fresh eval pass.