Google has launched Gemini 3.7 Flash, a new workhorse AI model for coding and AI agents, building on the success of its widely used Flash series.
Google is ramping up its launches, with the release coming just three weeks after Gemini 3.6 Flash. The 3.7 Flash is designed to deliver substantial improvements across software engineering, knowledge work and web development workflows, with an introductory price that is half the cost of 3.6 Flash per million tokens, FirstPost reported.
The new 3.7 Flash achieves strong first-pass code accuracy and offers improved performance when generating production-ready code. For web development, the model generates more functional layouts and feature-complete applications in fewer prompts.
For UI adherence, the model shows high design fidelity and parity with reference inputs, whether that is a screenshot, an image or a full design system. The model has also outperformed 3.6 Flash, scoring 1,588 in Arena AI’s WebDev Arena.
For knowledge-intensive fields such as finance, law and biosciences, 3.7 Flash offers stronger reasoning and accuracy, outperforming 3.6 Flash on complex document processing (34.0% vs 22.0% on GDP.pdf) and real-world business workflow automation (30.4% vs 17.0% on AutomationBench).
Google promises a noticeably improved developer experience over 3.6 Flash. This time, the model has been streamlined to follow instructions with greater fidelity, reason more carefully and put more effort into multistep planning and tool calls. Its more disciplined execution is designed to reduce manual oversight and the need for retries across engineering workflows.
3.7 Flash will be available at an introductory price through the end of the year, costing $0.75 per 1 million input tokens and $3.75 per 1 million output tokens. Combined with its improved performance, the pricing is designed to help developers and customers scale production-ready AI agents more cost-effectively.
Gemini 3.7 Flash is shipping with updated safeguards against misuse in chemical, biological, radiological and nuclear (CBRN) and cyber domains. At the same time, Google says it enables beneficial use cases in line with its approach to bioresilience and its cyber programme.
Gemini 3.7 Flash still shares some common limitations of foundation models, including occasional hallucinations, slow responses and timeout issues. Google says it has strengthened jailbreak protections and other safety measures. The model’s knowledge is current up to March 2026 for some domains, although certain areas may still rely on information available only up to January 2025.