Overview
- Google released Gemini 3.7 Flash on Thursday, August 13, 2026, and began rolling it out to Gemini Spark subscribers, the Gemini API, Google AI Studio, Android Studio, Google Antigravity and enterprise tools.
- DeepMind published benchmark gains showing large jumps in coding tests, including FrontierCode rising to 43.6% from 34.4% and DeepSWE jumping to 65.3% from 49.0%, and reported improvements on web development and workflow tests.
- Google is offering an introductory price through December 31, 2026 of $0.75 per 1M input tokens and $3.75 per 1M output tokens and has said those rates will increase to the prior levels on January 1, 2027, which matters for teams that run long agent loops.
- The model is natively multimodal with support for text, images, video, audio and PDFs, very large token limits, function calling and three thinking_level settings to trade off latency against deeper multi-step reasoning.
- The launch continues DeepMind’s rapid Flash cadence even as the flagship Gemini 3.5 Pro remains unreleased and leadership changes and new ultrafast offerings from rivals shift the market toward speed, cost and reliability, which will force developers to test migrations and plan for higher costs after the intro window.