Google ships Gemini 3.7 Flash three weeks after 3.6: the workhorse for coding and agents

Google released Gemini 3.7 Flash only three weeks after its predecessor 3.6, an update cadence that until recently would have sounded absurd for models of this weight. The official framing is "our most intelligent workhorse model for coding and agents": not the most powerful model in the lineup (that remains Pro), but the one built to grind through large volumes of daily work at the best quality-per-dollar.

The keyword is workhorse. In the Gemini family, Flash models sit in the middle band: smarter than the small Flash-Lite tier built for raw speed, far cheaper than the Pro tier dedicated to deep reasoning. It is the band used to build coding assistants, enterprise chatbots, and agents that carry out tasks autonomously, where per-request cost matters as much as quality.

The claimed improvements focus on programming and agentic workflows: on real-world software problem-solving tests the jump over 3.6 is substantial (from 49% to over 65% on one reference benchmark). Google pairs the launch with teaser pricing: $0.75 per million input tokens through year-end, set to double from January 2027.

For ordinary users the change arrives quietly inside Google products: the model already powers the Gemini experience for subscribers and the developer tools. The real story is the pace: three-week release cycles suggest the big labs are shifting from event-style launches to continuous updates, like regular software.