Google announced the launch of Gemini 3.7 Flash, the latest model in its Flash series, with a focus on programming, intelligent agents, and complex workflows. The company says the new model delivers major improvements in software engineering, web development, and knowledge-intensive tasks, arriving just three weeks after the launch of Gemini 3.6 Flash as a result of developer feedback and algorithmic innovations that Google said it would use in future models.
Gemini 3.7 Flash is available at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026. As of January 1, 2027, the price will increase to $1.50 per million input tokens and $7.50 per million output tokens.
Improvements in Programming and Web Development
According to Google, the new model advances on Gemini 3.6 Flash in programming tasks such as debugging and problem-solving, with higher first-attempt accuracy and a stronger ability to generate production-ready code. In tests cited by the company, Gemini 3.7 Flash scored 43.6% versus 34.4% for the previous model on FrontierCode 1.1 Main, and 65.3% versus 49.0% on DeepSWE v1.1.
In web development, the model can create more usable interfaces and fully featured applications using fewer prompts. User-interface generation tasks also showed stronger adherence to the reference design, whether the reference was a screenshot, an image, or a complete design system. The model scored 1588 points on the Elo scale in Arena.ai’s WebDev Arena, compared with 1538 points for Gemini 3.6 Flash.
Better Performance in Knowledge and Workflows
Google says Gemini 3.7 Flash delivers better reasoning and accuracy in knowledge-intensive fields such as finance, law, and life sciences. In the GDP.pdf test for processing complex documents, the model achieved 34.0% versus 22.0% for the previous model. It also scored 30.4% versus 17.0% on AutomationBench, an evaluation of the completion of real-world business tasks.
At the developer-experience level, the company says the model adapts better to obstacles, asks for clarification of intent when necessary, and follows instructions more accurately. It also makes greater efforts in multistep planning and tool calling, which may reduce the need for manual oversight and retries in engineering workflows.
Availability and Use in Gemini Spark
Gemini Spark, available to Google AI Pro and Ultra subscribers in more than 160 countries, has begun using Gemini 3.7 Flash. Google introduced Spark at its I/O conference as a personal artificial-intelligence agent that operates around the clock and carries out actions under the user’s direction. According to the company, the model update helps Spark perform knowledge work more efficiently, with improved use of Google Workspace application tools and more accurate outputs in complex, multiskill tasks.
Developers can try the model through Google Antigravity or the Gemini API via Google AI Studio and Android Studio. It is also available to organizations through the Gemini Enterprise Agent Platform and the Gemini Enterprise application. Individuals can access it through Spark inside the Gemini app if they are Google AI Pro and Ultra subscribers in supported countries.
Safety Controls
Gemini 3.7 Flash ships with updates to safeguards against misuse in chemical, biological, radiological, and nuclear areas, as well as cyber misuse, while Google continues to develop the coverage and robustness of these controls. The company referred users to the model card for more detailed information.