Gemini 3.7 Flash lands three weeks after 3.6
Google's workhorse model gains sharply on coding and agent benchmarks, with introductory pricing held to the end of the year.
2 minMain AI Hub
Google released Gemini 3.7 Flash on 13 August, three weeks after the 3.6 release it replaces. The company positions it as a workhorse for coding and agent work rather than a frontier reasoning model, and the published figures are consistent with that framing: the gains are concentrated in software engineering, interface generation and multi-step tool use.
On FrontierCode 1.1 Main the model scores 43.6% against 34.4% for 3.6 Flash. DeepSWE v1.1 moves from 49.0% to 65.3%. The WebDev Arena Elo rating rises from 1538 to 1588. Two benchmarks aimed at document handling and workflow automation show the largest relative movement: GDP.pdf from 22.0% to 34.0%, and AutomationBench from 17.0% to 30.4%.
Pricing is introductory and dated. Until 31 December 2026 input costs $0.75 per million tokens and output $3.75. From 1 January 2027 both double, to $1.50 and $7.50. Anyone budgeting a deployment on the current numbers should note that the second figure, not the first, is the standing price.
Access is through Google AI Studio, Android Studio and the Gemini Enterprise Agent Platform, with availability in Gemini Spark for Pro and Ultra subscribers across more than 160 countries.
The release cadence is itself the story for anyone maintaining an evaluation harness. A three-week gap between point releases in the same family leaves little room for the kind of independent measurement that would let a buyer verify these numbers before the next model arrives.
Retold from Google. This is a summary in our own words; follow the link for the original reporting.