Google has shipped another faster, cheaper Gemini Flash model. The catch: every improvement to the workhorse line makes the missing flagship harder to ignore.

The timeline is striking. In May, Google said Gemini 3.5 Pro would arrive in June; it did not. Three weeks after releasing Gemini 3.6 Flash, the company instead unveiled Gemini 3.7 Flash, positioning it as a more capable option for coding, knowledge work and agentic tasks. Google said the model “better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity.”¹

Google’s immediate pitch is speed and economics. The introductory rate is $0.75 per million input tokens and $3.75 per million output tokens—half the initial price of 3.6 Flash—and 3.7 is being rolled into the Gemini API, AI Studio and Gemini Spark for AI Pro and Ultra subscribers. CEO Sundar Pichai called Flash models “workhorses that offer performance at a great price,” arguing that rapid updates get gains into developers’ hands sooner.Sundar Pichai (sundarpichai)²

There is outside validation for that narrower mission. Perplexity CEO Aravind Srinivas said the Flash line works well as “fast, cost-efficient subagents” in a multi-model setup, including Perplexity’s Computer harness.Aravind Srinivas (AravSrinivas)³ That endorsement supports Google’s value proposition: Flash need not be the most powerful model to be useful at scale.

But the release also sharpens the opposing reading. Benchmark improvements are real, yet critics question whether a new Flash version after only three weeks signals a breakthrough or a substitute for the overdue Pro model. One account noted that Google’s latest announcement was “not the long-awaited 3.5 Pro,” and argued businesses relying on its AI tools must settle for an incrementally better Flash model for now.⁴

Google has not explained the fate of 3.5 Pro, while saying it is training Gemini 4. Until that ambiguity ends, the company’s bargain-priced progress will remain both an achievement and a distraction.