Google Launches Gemini 3.7 Flash for Coding and Agentic Work

What happened

Google has introduced Gemini 3.7 Flash, a new model positioned for software engineering, web development and knowledge-work tasks that require multiple steps. In its product announcement, Google describes the model as a fast, cost-conscious option for developers and teams building applications around coding and agentic workflows.

The company said Gemini 3.7 Flash is available through the Gemini API. Google listed introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens, with the stated promotional pricing running through December 31, 2026.

GitHub has also announced that Gemini 3.7 Flash is available in GitHub Copilot. The integration gives Copilot users a route to access the model within a coding environment rather than solely through Google's API.

Source evidence

Google's announcement is the primary source for the release, its intended use cases and its API pricing. The company specifically identifies software engineering, web development and agentic knowledge work as areas of focus for Gemini 3.7 Flash.

Google's published prices provide a concrete starting point for developers assessing the model's operating costs: $0.75 per million input tokens and $3.75 per million output tokens during the introductory period. The announcement sets an end date of December 31, 2026 for those rates.

GitHub's changelog independently confirms the availability of Gemini 3.7 Flash in GitHub Copilot. That announcement supports the conclusion that the model is being distributed through a major third-party developer product in addition to Google's own API offering.

Latent Space highlighted the release in its AI news coverage, while the underlying details on the model and pricing come from Google's announcement and GitHub's Copilot changelog.

Why it matters

Google is targeting a category of work where speed, cost and repeated model interactions can all affect whether a system is practical to deploy. Coding assistants, web-development tools and multi-step task systems may generate substantial input and output volumes, making token pricing relevant alongside model quality.

The listed API rates allow teams to make an initial comparison with alternative models. Still, token prices do not by themselves determine deployment costs. Prompt sizes, response lengths, retry rates, tool calls and the degree of human review needed for generated code can substantially change the total cost of a real application.

Availability in GitHub Copilot could broaden the model's exposure among developers. Rather than requiring a direct API integration, Copilot users may be able to evaluate Gemini 3.7 Flash in an established coding workflow, subject to the availability details and plan support set by GitHub.

What to watch

The central question will be how Gemini 3.7 Flash performs on practical development work. Developers will likely assess code correctness, debugging performance, reliability across programming languages, latency and its ability to work with the context of larger repositories.

GitHub's rollout details will also be important. Users will want clarity on which Copilot plans, regions and product interfaces include the model, as well as any usage limits that apply.

Finally, organizations considering a longer-term deployment will need to monitor pricing after the introductory period ends. Google's stated rates run through December 31, 2026, leaving future pricing as a factor in long-term cost planning.