Gemini 3.7 Flash lands: smarter coding workhorse, half the price
Google's cheapest capable model just jumped on coding, web, and agent benchmarks — and the $0.75/$3.75 intro rate holds only through December.
Copy markdown
The coding jump is real
Gemini 3.7 Flash hits 43.6% on FrontierCode 1.1 (up from 3.6 Flash's 34.4%) and 65.3% on DeepSWE v1.1 (from 49.0%), with a 1588 WebDev Arena Elo. At the Flash tier that means fewer retries and less babysitting inside your agent loops.
Half price — with a December clock
Intro pricing is $0.75 per 1M input and $3.75 per 1M output, half of 3.6 Flash — but only through Dec 31, 2026, then it doubles to $1.50/$7.50. It's the cheapest capable coding model for the rest of the year; budget for the January step-up.
Where to wire it in today
Live now as gemini-3.7-flash across the Gemini API, AI Studio, Android Studio, the Antigravity agent IDE, and Gemini Spark for AI Pro/Ultra in 160+ countries. No waitlist — repoint your Flash calls and go.
Not just code: docs and automation nearly doubled
Document comprehension jumped to 34.0% on GDP.pdf (from 22.0%) and workflow automation to 30.4% on AutomationBench (from 17.0%). That's a real unlock for PDF-heavy RAG and multi-step automation agents you'd have paid a frontier model to run.