Google has announced Gemini 3.7 Flash, a new AI model designed for software development, web development, and autonomous agent workflows. The company says it is the most capable Flash model to date for coding and agents, with improvements meant to help artificial intelligence take on complex tasks with less human intervention.
The launch is just three weeks after Gemini 3.6 Flash, highlighting the speed at which Google is iterating on its Gemini lineup. It’s not a general-purpose refresh. Gemini 3.7 Flash is more about practical, production-oriented work—debugging, fixing things, writing code, and walking through multiple steps. In web development, the model can also generate more functional layouts and even feature-complete applications from fewer prompts, while showing strong design adherence when given reference inputs like screenshots, images, or full design systems.
Google says the model produces better first-pass code and scores higher on software-engineering benchmarks. FrontierCode 1.1, Gemini 3.7 Flash, scored 43.6%, and Gemini 3.6 Flash scored 34.4%. Its DeepSWE v1.1 score was also improved from 49.0% to 65.3%.
Another big focus is web design. Google says Gemini 3.7 Flash can create more functional layouts and feature-complete applications with fewer prompts and also improve its ability to match designs provided through screenshots, images, or entire design systems. It scored 1,588 Elo on WebDev Arena, up from 1,538 for the last model.
Also Read: Terafab Chip Factory in Texas: Elon Musk’s Two Companies Set to Invest
The model also puts more emphasis on agentic workflows, allowing it to act on your behalf while still remaining under your direction. Google says it’s more adept at adjusting when an operation hits a snag, asking questions when it needs clarification and following instructions more accurately. The update also comes to Spark, making it more efficient for knowledge work by improving tool use across Google Workspace apps such as Docs, Sheets, and Gmail. Its improved multi-step planning and execution aim to reduce the number of failed loops, the need for manual supervision, and the number of repeated attempts.
The price might be as important as the performance improvements. Until Dec. 31, 2026, Google is offering its Gemini 3.7 Flash model at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens. That’s half the original pricing of Gemini 3.6 Flash, and Google is also applying the promotional rate to 3.6 Flash.
The model is generally available through the Gemini API and supports a 1-million-token context window, code execution, function calling, search grounding, and computer use in preview. Developers can also adjust its thinking level between low, medium, and high depending on the balance they need between speed, reasoning, and cost. For developers, that faster cadence creates its own operational question. Models can improve quickly, but production teams still have to benchmark new releases against their repositories, prompts, tool schemas, and failure modes before changing a deployment.
Google is also making Gemini 3.7 Flash the default model for its Antigravity agent and is bringing it to Gemini Spark for eligible subscribers. The blend of more robust execution and lower entry-level pricing could position the new Flash model as one of Google’s more important launches this year for developers working on coding tools and autonomous artificial intelligence systems.
