Google has launched Gemini 3.7 Flash, just three weeks after Gemini 3.6 Flash shipped, continuing the monthly cadence that now defines the Flash model line. The new model targets software engineering and automation workflows, posting benchmark improvements that go well beyond incremental.
Coding is where the numbers are sharpest, on the DeepSWE v1.1 benchmark, 3.7 Flash scores 65.3% against 49.0% for its predecessor, and FrontierCode 1.1 Main climbs from 34.4% to 43.6%. Google attributes both gains to improvements in multi-step planning and tool use, with the model described as better at adapting to roadblocks and following instructions with higher fidelity. In practice, that means fewer retries and less manual correction in engineering workflows.
Automation and Web Dev Gains
The AutomationBench score rises from 17.0% to 30.4% over 3.6 Flash, a jump that matters for developers relying on Gemini to handle real business workflows without constant intervention. Web development performance moved similarly: the Arena.ai WebDev market Elo score went from 1538 to 1588, and Google notes the model generates more functional layouts and feature-complete apps in fewer prompts.
Introductory Pricing at Half the Previous Launch Rate
Until the end of 2026, Gemini 3.7 Flash carries an introductory price of $0.75 per million input tokens and $3.75 per million output tokens, half the launch price of Gemini 3.6 Flash. The cut likely reflects Google’s confidence in volume growth, or at minimum its intent to keep developers on the Flash line rather than evaluating alternatives.
Availability covers Google AI Studio, Android Studio, the Gemini Enterprise Agent Platform, the Gemini Enterprise app, and Google Antigravity. In the Gemini consumer app, 3.7 Flash is rolling out to Spark, the 24/7 agent feature, with access requiring an AI Pro or Ultra subscription. Developers building on Apple’s local AI in iOS 26 are operating in a very different model-update environment by comparison. The release also is expected to ship with updated safeguards targeting CBRN and cyber offense misuse scenarios.
Google has been working to improve Gemini 3.5 Pro’s coding capabilities since at least June, updated its training data in an attempt to fix performance issues, and came away disappointed. The Flash line is now clearly where Google is shipping gains. With Gemini 4 not expected until November or December, three-week release cycles suggest Google would rather ship working improvements on Flash than wait for a Pro release that still isn’t ready.
