Google has launched Gemini 3.7 flash Google has launched Gemini 3.7 Flash, its latest AI model optimised specifically for complex software engineering tasks, web development and autonomous agent workflows. The newest model arrived three weeks after the release of Gemini 3.6 Flash, and the company says that it introduces substantial performance gains in multi-step planning, instruction following and code generation, while debuting at half the per-token introductory price of its predecessor.Major Gains in Coding, Web Design, and Document AnalysisAccording to benchmark results released by Google, Gemini 3.7 Flash delivers marked improvements over the 3.6 model across several technical domains. The company claims that the updated model achieved higher first-pass code accuracy and produced more production-ready code, scoring 43.6% on FrontierCode 1.1 (up from 34.4%) and 65.3% on DeepSWE v1.1 (up from 49.0%).It also said that in UI and web application design, the model demonstrated stronger layout functionality and reference-matching accuracy, earning an Elo score of 1588 compared to 3.6 Flash’s 1538 on Arena.ai’s WebDev Arena. For complex data processing, for fields like finance, law, and biosciences, Gemini 3.7 Flash significantly outperformed the previous version on document-processing evaluations and real-world business automation tasks.
Lower API pricing and Google Workspace upgrades
To boost enterprise adoption and scalable deployment of autonomous agents, Google is offering Gemini 3.7 Flash at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens through the end of the year. The company noted that the model is designed to “think more diligently,” requiring less manual developer oversight, fewer retries, and offering better execution during multi-step tool calls.Alongside developer availability via Google’s API channels, Gemini 3.7 Flash is being integrated immediately into Gemini Spark, Google’s 24/7 personal AI agent available to Google AI Pro and Ultra subscribers in over 160 countries. The update is expected to improve Spark’s execution across Google Workspace applications, enhancing accuracy in complex, multi-app workflows.