
Google launches Gemini 3.7 Flash, an improved version of its workhorse model for coding and agents, just three weeks after the release of 3.6. The new model delivers significant gains in software development, workflow automation, and document understanding — all at half the price through the end of 2026.
Higher performance, lower price
Gemini 3.7 Flash outperforms its predecessor on several key benchmarks. In software development, it scores 43.6% on FrontierCode 1.1 Main (vs. 34.4% for 3.6) and 65.3% on DeepSWE v1.1 (vs. 49%). On the web, its Elo score on WebDev Arena rises from 1,538 to 1,588, with better adherence to reference designs.

The improvements extend beyond coding. The model also advances on complex tasks such as document analysis (34% on GDP.pdf, up from 22%) or business workflow automation (30.4% on AutomationBench, up from 17%). Google highlights more rigorous execution: fewer backtracks, better handling of blockages, and greater adherence to instructions.
Fast deployment and aggressive pricing strategy
Google’s release pace is remarkable: just three weeks separate the 3.6 from the 3.7. This cadence is driven by developer feedback and algorithmic optimizations, according to the company. The model is available today via the Gemini API, AI Studio, Android Studio, or the Enterprise Agent Platform.
Until December 31, 2026, pricing is set at $0.75/million input tokens and $3.75/million output tokens — half the launch price of the 3.6. Starting January 1, 2027, prices will double to $1.50 and $7.50. A strategy to encourage businesses to adopt the model before the increase.
Gemini Spark upgrades to 3.7, but not for everyone
Google AI Pro and Ultra subscribers now get 3.7 Flash in Gemini Spark, the personal AI agent launched at I/O. Spark gains efficiency in knowledge tasks and Google Workspace tool usage, such as file consolidation or email drafting.
However, the public interface of Gemini (outside Spark) continues to use 3.6 Flash for now. This distinction reflects Google’s intent to reserve the latest advancements for its premium subscribers.
Security and context: a subtle but strategic update
Gemini 3.7 Flash integrates stronger safeguards against malicious use in CBRN (chemical, biological, radiological, nuclear) and cyber-offensive domains, while allowing legitimate use cases. An update aligned with Google’s bioresilience and cybersecurity principles.
This release comes amid internal reshuffling at Google DeepMind, with notable departures (Jeff Dean, Oriol Vinyals) and a governance reorganization. The giant remains competitive on benchmarks but struggles to deliver its flagship model, 3.5 Pro, initially slated for June 2026.
With 3.7 Flash, Google confirms its strategy: rapidly iterating on its "workhorse" models to close the gap with competitors, while leveraging pricing. It remains to be seen if this approach will convince developers and businesses, as expectations grow for 3.5 Pro and, above all, the future Gemini 4.
Sources
Google — The Keyword — Introducing Gemini 3.7 Flash
Android Authority — Gemini 3.7 Flash is here with better coding, reasoning, and more
VentureBeat — Google’s Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut
Ars Technica — Google announces Gemini 3.7 Flash just three weeks after previous release
9to5Google — Gemini 3.7 Flash launches three weeks after last model, live in Spark





Comments 0
Sign in to join the discussion.
Be the first to comment.