Google isn’t waiting around. Three weeks after shipping Gemini 3.6 Flash, the company has already dropped its successor — Gemini 3.7 Flash — and the pricing story is as aggressive as the release cadence. Launch costs sit at $0.75 per million input tokens and $3.75 per million output tokens, a full 50 percent cheaper than what 3.6 Flash commanded at its debut just last month. Both models now share identical pricing through year’s end, though Google’s phrasing suggests that’s a temporary arrangement.
The real headline is in the benchmarks. On FrontierCode, Gemini 3.7 Flash scores 43.6 percent, up from 34.4 percent for its predecessor — a jump of more than nine points. DeepSWE tells a similar story: 65.3 percent versus 49.0 percent. Google claims those numbers put the model ahead of Anthropic’s Claude Sonnet 5 and OpenAI’s GPT-5.6 Terra, positioning 3.7 Flash as a legitimate threat in the coding agent space, not just a budget option.
Google attributes the rapid gains to what it calls “awesome algorithmic improvements,” which is the kind of vague phrasing that would normally raise eyebrows. But the benchmark deltas are concrete enough to warrant attention. Web development, document comprehension, and business process automation all show gains too, suggesting the improvements aren’t narrowly tuned to a single benchmark suite.
The model is available now through Google’s API, AI Studio, and Antigravity. The speed here is notable — three weeks between Flash iterations is unprecedented, even by Google’s already frantic release standards. It signals that whatever pipeline produced these gains is still producing, and that the Flash line has become Google’s primary vehicle for shipping rapid improvements. The question is whether competitors can match this cadence without sacrificing the safety evaluations that typically slow down model releases. Google clearly thinks the tradeoff is worth it.
