Gemini 3.7 Flash posts large coding gains and undercuts its own predecessor by half
Released 13 August, three weeks after 3.6 Flash. FrontierCode 1.1 Main rises to 43.6% from 34.4%; DeepSWE v1.1 to 65.3% from 49.0%; AutomationBench to 30.4% from 17.0%. Introductory pricing is $0.75/$3.75 per million tokens through 2026, half the previous Flash cost.
Three weeks between releases and a 16-point jump on DeepSWE is a fast iteration loop by any standard. Pairing it with a 50% price cut on the same tier says Google has decided the coding and agent workload is worth buying rather than earning gradually.
The benchmark movement is broad rather than isolated, which is what makes it credible: FrontierCode 1.1 Main from 34.4% to 43.6% on production code quality, DeepSWE v1.1 from 49.0% to 65.3%, AutomationBench from 17.0% to 30.4%, and Code Arena web development Elo from 1538 to 1588. Improvements concentrated in one benchmark usually mean targeted optimisation; improvements across four suggest something real in the training.
The pricing has an expiry attached and it is stated up front. $0.75 in and $3.75 out per million tokens through the end of 2026, then $1.50 and $7.50 from 1 January 2027 — a doubling, announced at launch.
That is the second frontier lab in a week to make a promotional price a headline feature. It is worth remembering how the last one went: Anthropic announced a Sonnet 5 increase for September and then cancelled it under competitive pressure. Announced future prices in this market have not been reliable predictors of actual future prices.
Google — Gemini 3.7 Flash: our most intelligent workhorse model → · VentureBeat — Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut → · TechTimes — Google Launches Gemini 3.7 Flash: Coding Gains at Half-Price Through Year-End →