Gemini 3.7 Flash launches three weeks after last model, live in Spark

Three weeks following Gemini 3.6 Flash, Google has unveiled Gemini 3.7 Flash, showcasing the tech giant's rapid release cycle. The new model's enhancements stem from developer input and algorithmic innovations, aiming to deliver better performance across various domains, such as software engineering, web development, and knowledge work.

Google highlights substantial improvements in coding, noting significant gains in debugging and issue resolution for Gemini 3.7 Flash over its predecessor. Benchmarks like DeepSWE v1.1 and FrontierCode 1.1 Main demonstrate the model's increased performance, showcasing its ability to generate functional layouts and feature-complete apps with fewer prompts, outperforming the previous model's Elo score of 1538 on Arena.ai's WebDev Arena.

The improvements extend to UI generation, with Gemini 3.7 Flash exhibiting high design adherence and parity based on reference inputs, be it screenshots, images, or full design systems. Additionally, in knowledge-dense fields like finance, law, and biosciences, the model demonstrates enhanced reasoning and accuracy.

Developers can expect Gemini 3.7 Flash to better adapt to challenges, clarify intent, and follow instructions with greater fidelity. The model's more disciplined approach to multi-step planning and tool calls results in reduced manual oversight and fewer retries during engineering workflows.

Safety is also a priority, with updated safeguards against misuse in areas like CBRN and cyber offense, while enabling beneficial applications in alignment with Google's bioresilience and cyber programs. During its initial launch, Gemini 3.7 Flash is available at a discounted price of $0.75/1M input tokens and $3.75/1M output tokens, half the cost of the previous model at launch.

The Gemini app will roll out Gemini 3.7 Flash to Spark users (AI Pro and Ultra subscription required) in the coming weeks, promising a more efficient personal agent for knowledge work with improved tool use for Google Workspace apps. The enhanced capabilities will also be available in Google Antigravity, AI Studio, Android Studio, Gemini Enterprise Agent Platform, and the Gemini Enterprise app.