Google Rolls Out Gemini 3.7 Flash for Coding and AI Agents

Google has released Gemini 3.7 Flash, a new language model optimized for coding and AI agent work. The company is positioning it as a cheaper option for enterprises building autonomous systems capable of planning tasks, using developer tools, and executing multi-step processes without constant human oversight.

What the model shows

According to Google, the new model improved results on coding tasks involving debugging, resolving reported issues, and generating code ready for production deployment. The company also highlights better handling of multi-step planning and tool calls, which matters for agents carrying out complex tasks without continuous human supervision.

Our most intelligent model yet for everyday coding and agent work - Google

It handles obstacles better, clarifies intent when needed, and follows instructions more precisely. It puts more effort into multi-step planning and tool calls - Google

The math for businesses

Pricing is the key element of this launch. Google is offering Gemini 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens, half of what Gemini 3.6 Flash cost. The introductory rate holds through the end of 2026, after which it rises to $1.50 and $7.50 per million tokens.

For companies scaling production agents that process millions of queries per month, a cut like this translates directly into operating costs. Google is explicit that the lower price is meant to encourage business customers to build agentic systems at larger scale, rather than limiting themselves to single chat queries.

Three weeks after its predecessor

Gemini 3.7 Flash arrived just three weeks after Gemini 3.6 Flash, which Google attributes to rapid incorporation of developer feedback and algorithmic improvements that the company says it will carry forward into future models. Such a short gap between consecutive Flash releases stands in contrast to the flagship Gemini 3.5 Pro, which remains in partner testing with no confirmed launch date.

That contrast matters to investors watching whether Google DeepMind is keeping pace with Anthropic and OpenAI in the race for the best coding models. Regular, cheaper releases in the Flash line let the company maintain a presence in the developer tools market even as the flagship model's launch drags on.

Integration with developer tools

The model has been added to Google Antigravity, Google's platform for agent-first work spanning frontend, full-stack, and scientific use cases. At the same time, Gemini 3.7 Flash was immediately made available in Gemini Spark, Google's subscription agentic service for AI Pro and Ultra plan customers in more than 160 countries.

This simultaneous rollout across several products shows that Google treats Flash as a workhorse model for a broad tool ecosystem, not just a one-off release for lab testing.

What it means for Polish companies

For Polish development teams and companies deploying AI agents to automate processes, the 50 percent price cut represents a real opportunity to scale projects built on Google's API more cheaply. It's worth noting that the introductory rate expires at the end of 2026, so longer-horizon budget plans should account for costs doubling starting in January 2027.

Companies comparing model providers for coding tasks gain another reference point in benchmarks such as DeepSWE and FrontierCode, which Google published alongside the launch.