I look at the price first, the tables second. $0.75 per million input tokens, $3.75 per million out. Half of what the previous Flash launched with a few months ago.
Fact
The facts: on 13 August 2026 Google announced Gemini 3.7 Flash. Introductory pricing is $0.75 per 1M input tokens and $3.75 per 1M output, in force until 31 December 2026. By Google's numbers the model lifts FrontierCode 1.1 from 34.4% to 43.6%, DeepSWE v1.1 from 49.0% to 65.3%, and WebDev Arena Elo from 1538 to 1588. It is available in the Gemini API, Google AI Studio, Antigravity and the Gemini Enterprise Agent Platform, and through Gemini Spark for subscribers in over 160 countries.
The numbers are theirs. Every company arranges its benchmarks to look good, and that is not an accusation, that is the craft. But the direction is clear without any table: the workhorse you run all day inside an agent got twice as cheap while getting better.
The expensive model wins the demo. The cheap one decides who ships a product at all.
That price is introductory and it expires on 31 December. If you are costing a project that starts next year, run the numbers with both figures, not just the low one. I have seen what happens when someone builds their economics on a promotion.
Official primary sources
Google Blog - Introducing Gemini 3.7 FlashThe visual is generated code-art, no third-party images.