Gemini 3.7 Flash halves the price, and the discount has an expiry date
Released 13 August at $0.75 per million input tokens and $3.75 output — half of 3.6 Flash's launch price. Standard rates of $1.50 and $7.50 begin on 1 January 2027. The coding and agent benchmarks moved a long way.
Gemini 3.7 Flash arrived on 13 August, three weeks after 3.6 Flash, keeping the million-token context window and multimodal input while posting large gains where the money currently is: coding and agents.
The benchmark deltas are unusually wide for a point release. DeepSWE v1.1 went from 49.0% to 65.3%. AutomationBench went from 17.0% to 30.4% — not an improvement so much as a doubling. WebDev Arena Elo moved 1538 to 1588. Specs stay at 1M input, 64k output, with function calling, search as a tool and computer use.
The pricing is the part to read carefully. Introductory rates are $0.75 per million input tokens and $3.75 output, half what 3.6 Flash launched at. Standard rates of $1.50 and $7.50 start on 1 January 2027. That is not a price cut. It is a price cut with a published end date, which is a different instrument.
An introductory rate that expires on a known day is designed to move integrations now and collect later, and it works because switching costs accrue quietly. A team that builds against Flash in September is not going to re-plumb in January over a doubling they agreed to in advance.
It is also the second pricing structure this month that makes the headline per-token number insufficient on its own. With eighteen model releases from fifteen providers in a single month, comparing models on a single figure is becoming a category error.
VentureBeat — Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut → · DataCamp — Gemini 3.7 Flash: Features, Benchmarks, and Pricing → · Memeburn — Google Launches Gemini 3.7 Flash but Its Low Price Has an Expiry Date →