Google DeepMind introduces Gemini 3.7 Flash for coding and agent workflows

Gemini 3.7 Flash becomes Google's new workhorse model

Google DeepMind published the Gemini 3.7 Flash model card on August 13, 2026. The model is the next iteration of the Gemini 3 Flash line and is positioned for coding, agentic workflows and enterprise automation.

Gemini 3.7 Flash supports text, images, audio and video as inputs, with a context window of up to one million tokens and text outputs up to 64K tokens. Google says developers can tune thinking behavior to balance quality, cost and latency.

Distribution and safety

DeepMind lists Gemini 3.7 Flash across the Gemini app, Gemini Enterprise products, Google AI Studio, the Gemini API and Google Antigravity. Its evaluation set covers reasoning, coding, long-context work, multimodal tasks and agentic tool use.

The model card also documents limitations such as hallucinations and occasional latency or timeout issues. DeepMind reports that Gemini 3.7 Flash did not reach the tracked or critical capability levels in its Frontier Safety Framework, although the company continues to deploy safeguards for areas including cyber offense and CBRN misuse.

For developers, the practical takeaway is a faster model tier designed to carry more of the coding and agent workload while preserving explicit controls over reasoning cost and latency.