Google is moving fast on its cheaper AI workhorse while leaving its more ambitious model conspicuously offstage. Gemini 3.7 Flash may be a useful update; it is also a reminder that Gemini 3.5 Pro has yet to arrive.

Three weeks after releasing Gemini 3.6 Flash, Google rolled out 3.7 Flash, saying it improves coding, web development and knowledge-work tasks. The company says the model handles multi-step work more carefully: “It better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity.”¹

The launch is paired with an aggressive introductory rate — $0.75 per million input tokens and $3.75 per million output tokens, half the original 3.6 Flash price.¹ Google has also put the new model behind Gemini Spark, its agent for AI Pro and Ultra subscribers, framing the update as stronger tool use across Workspace workflows.²

Google’s leadership presented the rapid cadence as the point, not a distraction. Sundar Pichai called Flash models “workhorses that offer performance at a great price” and said 3.7 delivered significant gains in coding and agentic work just weeks after 3.6.Sundar Pichai (sundarpichai)³ DeepMind chief Demis Hassabis similarly said the update brought “major upgrades” and an introductory price at half the prior cost.Demis Hassabis (demishassabis)⁴

There is outside validation for the value proposition. Perplexity chief executive Aravind Srinivas said his company uses Flash models as “fast, cost-efficient subagents” in its Computer harness.Aravind Srinivas (AravSrinivas)⁵

But the rollout has sharpened the counterargument. Google promised Gemini 3.5 Pro for June and has not shipped it; meanwhile, 3.7 Flash’s benchmark gains and discount may keep developers in the ecosystem without resolving concerns that Google is trailing OpenAI and Anthropic at the frontier.⁶ The company has declined to discuss 3.5 Pro’s fate while signaling optimism about Gemini 4 — leaving open the possibility that the missing Pro model is being skipped rather than delayed.¹