Google he announced Gemini 3.7 Flasha new iteration of its family of multimodal templates optimized for programming, web development and intelligent agent management. Just twenty-one days after the release of the previous version, the Mountain View company has accelerated the pace of distributions, introducing an architecture capable of processing a maximum input context equal to 1,048,576 tokens and up to 65,536 tokens outgoing.
The update aims to reduce operating costs for developers and companies, setting a promotional list price valid until the end of 2026 equal to $0.75 per million token input And $3.75 per million token output. On benchmarks dedicated to software engineering, Gemini 3.7 Flash shows a clear performance increase compared to its predecessor 3.6 Flash: in the test DeepSWE v1.1 the problem resolution rate rises from 49.0% to 65.3%while the evaluation FrontierCode 1.1 Main goes from 34.4% to 43.6%.
In terms of the development of web interfaces and applications, the metrics measured on WebDev Arena see the Elo score grow up to 1588 pointsagainst the 1538 recorded by the previous release. Capabilities for generating functional layouts from screens, images or complete design systems benefit from greater adherence to the instructions given in the prompts. To consult the technical data sheets and official development documentation, refer to the portals provided by the manufacturer.
Gemini 3.7 Official Flash 21 days after the launch of 3.6
The algorithmic refinement also significantly affected document processing and the execution of multi-step business workflows. In tests GDP.pdfdesigned to measure information extraction from complex files, Gemini 3.7 Flash reaches a score of 34.0%exceeding the 22.0% achieved by version 3.6. Automated operational flows monitored on AutomationBench recorded a change from 17.0% to 30.4%highlighting superior discipline in planning and interacting with external tools.
The model natively supports textual, visual, audio, video and PDF format inputand therefore allows developers to configure the depth of the reasoning process (settable to low, medium and high levels). Google has updated its security protocols against the risks of abuse in the chemical, biological, radiological, nuclear and cyber attack sectors. Access to Gemini 3.7 Flash it is already available to subscribers of the AI Pro and Ultra plans within the Spark environment, as well as being integrated into Google Antigravity, AI Studio, Android Studio and the Gemini Enterprise platform.
