Google has launched Gemini 3.7 Flash, its latest cost-focused artificial intelligence model built around software development, autonomous AI agents and multi-step business workflows. The August 13 release arrives only three weeks after Gemini 3.6 Flash, highlighting the speed at which Google is updating its AI lineup as competition with OpenAI and Anthropic intensifies.

For developers, two changes stand out: improved coding performance and lower introductory pricing. Google is offering Gemini 3.7 Flash for $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026, half the original cost of Gemini 3.6 Flash.

Google is positioning the model as a lower-cost option for businesses building AI systems that can plan tasks, use software tools and complete multiple steps with less human intervention. That fits the company’s broader work on Gemini agents, developer tools and AI infrastructure.

Gemini 3.7 Flash makes a bigger coding push

Google says Gemini 3.7 Flash improves debugging, software issue resolution and production-ready code generation. Its published benchmarks indicate meaningful gains over the model it replaces.

On FrontierCode 1.1 Main, Gemini 3.7 Flash scored 43.6%, up from 34.4% for Gemini 3.6 Flash. On DeepSWE v1.1, the new model reached 65.3%, compared with 49.0% previously.

Web development also improved. Gemini 3.7 Flash recorded a WebDev Arena Elo score of 1,588 versus 1,538 for 3.6 Flash. Google says it can build more complete applications with fewer prompts and reproduce interfaces using screenshots, images or broader design systems as references.

Benchmarks do not guarantee identical results in real applications. Performance can vary depending on prompts, programming languages, tools and the complexity of a company’s workload, making independent testing important before production deployment.

AI agents could be the more important upgrade

Gemini 3.7 Flash is designed to do more than answer questions or generate isolated pieces of code. Google says the model is better at multi-step planning, calling tools and adapting when it encounters roadblocks.

On AutomationBench, which evaluates business workflows, Gemini 3.7 Flash scored 30.4%, compared with 17.0% for Gemini 3.6 Flash. Better reliability across longer workflows could reduce retries and human supervision when companies deploy AI agents.

This is becoming a major competitive area. Anthropic has similarly focused on coding and autonomous workflows, making advanced coding and autonomous AI agents an increasingly important battleground among AI developers.

Google also reports improvements outside programming. On GDP.pdf, an evaluation involving complex documents, Gemini 3.7 Flash scored 34.0%, up from 22.0% for 3.6 Flash. The company highlights potential uses in knowledge-heavy fields including finance, law and biosciences.

50% introductory pricing comes with an expiry date

Gemini 3.7 Flash’s pricing could be as important to businesses as its benchmark gains. Standard paid API access costs $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026.

From January 1, 2027, Google’s listed standard rates rise to $1.50 per million input tokens and $7.50 per million output tokens. Batch and Flex processing starts at $0.375 per million input tokens and $1.875 per million output tokens during the introductory period before also doubling in 2027.

Businesses estimating long-term costs should therefore not assume the 50% reduction is permanent. Developers can check current rates and conditions on the official Gemini API pricing page.

The economics matter because high-volume AI applications can process millions or billions of tokens. Businesses must consider not only token prices but also reliability, latency, retries, caching, grounding and the amount of human intervention a model requires.

Gemini Spark gets the new model immediately

Google is also rolling Gemini 3.7 Flash into Gemini Spark, its subscription-based AI agent service for Google AI Pro and Ultra customers in more than 160 countries. That gives Google a direct route for bringing the model’s agent capabilities to users beyond developers working with APIs.

The launch comes during a period of significant change around Google’s AI operation. Google co-founder Sergey Brin has been pushing key AI staff to intensify work around Gemini as Alphabet tries to keep pace with its strongest competitors.

Google DeepMind has also undergone a leadership overhaul, with Demis Hassabis stepping aside as chief in favour of his deputy, Koray Kavukcuoglu. Two original technical co-leads of Gemini have also left to co-found a startup.

The wait for Gemini 3.5 Pro continues

One major question remains unanswered. Google previously said Gemini 3.5 Pro was being tested with partners and was coming soon, but the company has not provided a firm release date.

That distinction matters because Flash is designed around a balance of performance, speed and cost, while Google’s flagship Pro model will face greater scrutiny over advanced reasoning and its ability to compete with leading models from OpenAI and Anthropic.

Gemini 3.7 Flash therefore represents more than another model update. Google is competing on coding quality, agent reliability and the economics of running AI at scale. Whether the benchmark gains translate consistently into production workloads will now be tested by the developers and businesses Google is trying to win over.