Three weeks. That's the whole gap.
Google shipped Gemini 3.6 Flash on July 21. On August 13 it shipped 3.7 Flash. Twenty-three days later. A year ago that cadence would have sounded like a typo. The workhorse tier is now on a near-monthly clock, and the reason is blunt: developers asked for better coding and agent behavior, and Google decided not to wait a quarter to deliver it.
Here's the scenario. You run a small consultancy. You wire Gemini into client pipelines to draft code and run multi-step agents. Your API bill is something you actually look at. Last month you were on 3.6. Now 3.7 is out and it's cheaper. Do you migrate? In this case, almost certainly yes.
The price dropped, with a clock attached
Input is $0.75 per million tokens, output $3.75. A million tokens is roughly 750,000 Chinese characters, or about three full novels. So generating three novels' worth of text costs you under thirty yuan. That's half of what 3.6 charged at launch.
The catch most summaries skip: this is an introductory rate that expires December 31, 2026. From January 1, 2027 it doubles to $1.50 / $7.50. Cheap now, but budget the renewal.
Real coding gains, not a miracle
Google's own numbers: DeepSWE v1.1 (long-horizon software engineering) goes from 49.0% to 65.3%; FrontierCode 1.1 Main (production code quality) from 34.4% to 43.6%; WebDev Arena Elo from 1538 to 1588. Every line is up. None is a blowout.
Honest read: on AutomationBench, the private business-workflow test, 3.7 scores 30.4%. That means it still fumbles roughly seven of ten multi-step automation tasks. Don't hand it your entire back-office flow and walk away.
Where it runs, and who's locked out
The developer channel opened the same day: Google AI Studio, Android Studio, and the Gemini Enterprise platform all serve it. Regular users reach 3.7 through the Spark agent inside the Gemini app, which requires an AI Pro or Ultra subscription. One cold but real caveat: users in the European Economic Area, the UK, Switzerland, and Nigeria can't pull 3.7 from the app yet, though the API is unaffected. Teams shipping to the EU should confirm region before deploying.
Who should switch, who should wait
If you write code, run agents, or chain multi-step tasks, 3.7 beats 3.6 clearly, and switching costs nothing today because the two share a price. But stack it against DeepSeek V4-Flash at $0.14 / $0.28 and Google still looks expensive. High volume on a tight budget? The cheaper endpoint wins.
Why this matters to you
If Gemini is already in your daily loop for writing, coding, or web builds, moving to 3.7 is essentially free upside: same price, better behavior, full 1M-token context. Just don't lock your annual cost model to today's promo rate. Use it while it's cheap, and plan for the January reset.
Related reading: Gemini, DeepSeek