Google announces Gemini 3.7 Flash just three weeks after previous release
Ars Technica Ryan Whitwam ● Covered by 2 sources
Google just pushed out Gemini 3.7 Flash, three weeks after 3.6 Flash. It’s meant to be cheaper and better at coding and business tasks.
Based on reporting by Ars Technica, Ryan Whitwam — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google has a new Gemini model out today, and it isn’t the long-rumored 3.5 Pro. The company is rolling out Gemini 3.7 Flash to replace 3.6 Flash, which itself arrived only three weeks ago.
Google is pitching 3.7 Flash as its workhorse model, the result of core optimizations and developer feedback. The big promise is better coding and stronger agentic performance, plus a lower introductory price to blunt the appeal of cheaper competing models.
Tulsee Doshi, a senior director at Google, says the new Flash model is clearly better at coding than the last one. On FrontierCode 1.1 Main, she points to a rise from 34.4 percent to 43.6 percent. On DeepSWE v1.1, the score moved from 49 percent to 65.3 percent.
The gains are not just for code. Gemini 3.7 Flash’s WebDev Arena score is up to 1,588 from 1,538, while GDP.pdf, which measures how well a model handles complex documents, climbs to 34 percent from 22 percent. AutomationBench, which checks common business workflows, rises to 30.4 percent from 17 percent.
So the pattern is familiar: another faster-than-expected model refresh, this time framed around practical improvements rather than a giant leap. Google seems more interested in keeping Flash sharp and competitively priced than waiting around for one neat flagship moment.
My take — AI-written commentary, not fact-checked reporting
This is the AI industry’s favorite trick now: release, benchmark, repeat, before anyone has finished memorizing the model name. It’s less grand strategy than treadmill management, with pricing and coding scores doing the heavy lifting. Fine for developers, mildly exhausting for everyone else.
Read more about this at: Ars Technica
