Google ships Gemini 3.7 Flash for coding and agents

Google released Gemini 3.7 Flash on August 13, calling it its “most intelligent workhorse model yet for coding and agents.” Against Gemini 3.6 Flash, the company reports production code quality up to 43.6% on FrontierCode 1.1 from 34.4%, DeepSWE v1.1 up to 65.3% from 49.0%, a WebDev Arena Elo of 1588 versus 1538, and AutomationBench (business-workflow completion) up to 30.4% from 17.0%.

Introductory pricing through December 31, 2026 is $0.75 per million input tokens and $3.75 per million output, roughly half of 3.6 Flash’s rate; standard pricing of $1.50/$7.50 kicks in January 1, 2027. It’s available now through Google Antigravity, AI Studio, and Android Studio for developers, the Gemini Enterprise Agent Platform for businesses, and Gemini Spark for AI Pro and Ultra subscribers in more than 160 countries.

What it means for you

The AutomationBench and DeepSWE jumps matter more than the headline Elo number if you’re using Gemini for anything agentic, business-workflow completion nearly doubling and coding benchmarks up double digits is a real capability gain, not just a price cut. If Flash-tier models are already in your routing mix, this is worth benchmarking against your own workload before the introductory pricing window closes at the end of the year, the same discipline behind picking the right AI tools for a small operation rather than defaulting to whatever tier is already wired in.