Google Launches Gemini 3.7 Flash — Cheaper Coding and Agent Model
Google announced Gemini 3.7 Flash on August 13, calling it its most intelligent workhorse model yet for coding and agents, with stronger benchmarks and an introductory price half of 3.6 Flash.
What happened
On August 13, Google introduced Gemini 3.7 Flash, which it calls its most intelligent workhorse model yet for coding and agents. The release arrives just three weeks after Gemini 3.6 Flash and, per the Gemini team, comes directly out of developer feedback and algorithmic work.
Benchmarks and price
Gains over 3.6 Flash are broad: FrontierCode 1.1 Main rises to 43.6% from 34.4%, DeepSWE v1.1 to 65.3% from 49.0%, and GDP.pdf, which tests work with complex documents, to 34.0% from 22.0%. AutomationBench, measuring real business workflows, scores 30.4% versus 17.0%. On Arena.ai's WebDev Arena, Elo reaches 1588, up from 1538.
The introductory price is half the original 3.6 Flash cost per million tokens — a clear sign that competition among fast, cheap models is now a price fight as much as a capability fight.
Why it matters
The demo material leans on agentic workflows: sub-agent orchestration for landing pages, robotic training loops that use multimodal understanding, and converting PDFs into interactive data stories. For teams building on top of models, cheaper agents mean more automation that can run in production.