Google released Gemini 3.7 Flash on August 13, 2026, aiming to give software teams a faster, cheaper model for coding and agent workflows. You can already try it via the Gemini API and Google AI Studio, with broad support across developer tools.
Where you can find it
Google is rolling out the model across multiple development environments:
- Google services: Available in Google AI Studio, the Gemini API, Android Studio, Antigravity, Spark, and Gemini Enterprise platforms.
- GitHub Copilot: Rolling out to Copilot Pro, Pro+, Max, Business, and Enterprise plans, as announced on the GitHub Changelog.
- Supported IDEs: You can select it in Visual Studio Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, the Copilot app, and the Copilot cloud agent.
For GitHub Business and Enterprise workspaces, admins must first turn on the Gemini 3.7 Flash Preview policy.
The numbers Google reported
Google positions this release as a clear step up from Gemini 3.6 Flash.
According to Google’s published benchmarks:
- DeepSWE v1.1: Rose from 49.0% on 3.6 Flash to 65.3% on 3.7 Flash.
- FrontierCode 1.1 Main: Increased from 34.4% to 43.6%.
- WebDev Arena: Scored 1588, up from 1538.
- GDP.pdf: Jumped from 22.0% to 34.0%.
- AutomationBench: Reached 30.4%, up from 17.0%.
Pricing gets a temporary cut. Through December 31, 2026, Gemini API access costs $0.75 per 1 million input tokens and $3.75 per 1 million output tokens. GitHub Copilot uses its own usage-based billing rules.
The part you shouldn’t trust yet
All benchmark scores and speed claims come straight from Google’s internal tests — nobody outside the company has verified them yet. Real projects involve complex tool loops and messy codebases. A benchmark score does not guarantee lower token usage or fewer bugs in your private repository. Also, Google’s introductory API price doubles on January 1, 2027, rising to $1.50 per 1M input tokens and $7.50 per 1M output tokens.
How to test it yourself
Do not switch your production tools immediately. Test it on real tasks first.
Pick a typical coding task with tool calls. Run it ten times on Gemini 3.6 Flash and ten times on Gemini 3.7 Flash using the same prompt. Track task success, how many retries it took, and total tokens spent. If 3.7 finishes tasks in fewer steps, it will save you money during the discount period. If it fails or loops often, the temporary discount will not help you.