Gemini 3.7 Flash ships: half the cost, big coding-agent gains
Google's new workhorse is live today across API, AI Studio and Vertex — and it arrived before the long-promised 3.5 Pro, which may be skipped for Gemini 4.
Copy markdown
Half the price, from launch day
Introductory pricing runs $0.75 per 1M input tokens and $3.75 per 1M output — literally half of 3.6 Flash — through Dec 31, 2026. Standard rates ($1.50/$7.50) kick in Jan 1, 2027.
Coding and agents jump hard
DeepSWE v1.1 climbs to 65.3% (from 49.0%), FrontierCode 1.1 to 43.6% (from 34.4%), and WebDev Arena Elo to 1588 (from 1538). Google tuned it for multi-step agent workflows with fewer retries and less hand-holding.
Live everywhere you build, today
Ship with it now via the Gemini API, Google AI Studio, Vertex AI, Android Studio and Antigravity, plus the Gemini app for Pro/Ultra subscribers in 160+ countries. It's the default workhorse — wire it into your agents.
Sharper on documents and knowledge work
Doc-parsing on the GDP.pdf benchmark rises to 34.0% (from 22.0%) and AutomationBench to 30.4% (from 17.0%), with gains in finance, law and biosciences reasoning — useful if you build over messy PDFs.
3.5 Pro slips — maybe for good
The bigger Gemini 3.5 Pro that was promised still hasn't shipped and has no timeline. Google now says it's “training for Gemini 4 and excited by early results,” hinting Pro may be skipped entirely.