Gemini 3.7 Flash is Google’s new AI workhorse
Google’s latest Flash model focuses on faster coding, stronger reasoning and lower costs for agentic workflows.
Google has introduced Gemini 3.7 Flash, positioning the model as a high-performance workhorse for coding, AI agents and knowledge-heavy tasks. Announced on 13 August 2026, the model arrives only three weeks after Gemini 3.6 Flash and brings improvements based on developer feedback and changes to its underlying algorithms.
Google is also launching Gemini 3.7 Flash at roughly half the original price of Gemini 3.6 Flash, making cost a key part of its appeal for developers running AI applications at scale.
Stronger coding and web development
Gemini 3.7 Flash is designed to improve how AI handles software engineering tasks, including debugging, resolving issues and producing usable code on the first attempt. On the FrontierCode 1.1 Main benchmark, the model scored 43.6%, compared with 34.4% for Gemini 3.6 Flash.
On DeepSWE v1.1, it scored 65.3%, up from 49.0%. The improvements are also aimed at web development. Google says the model can create more functional layouts and feature-complete applications with fewer prompts. It also follows design references more closely, with its WebDev Arena Elo rating increasing to 1588 from 1538.
Better at documents and multi-step tasks
The improvements extend beyond coding. Gemini 3.7 Flash also performs better on tasks involving large or complex documents and real-world workflows. On the GDP.pdf evaluation, which tests complex document understanding, the model scored 34.0%, compared with 22.0% for Gemini 3.6 Flash.
On AutomationBench, which measures the completion of practical workflows, it achieved 30.4%, compared with 17.0% for its predecessor. These capabilities could be useful for industries that work with large amounts of structured and unstructured information, including finance, law and biosciences.
Designed for AI agents
Google is also positioning Gemini 3.7 Flash as a model for agentic AI. These systems can plan tasks, use software tools and complete several steps with less human intervention. The model focuses on clearer intent handling, stronger instruction following and more reliable planning.
For developers, this could mean fewer retries and less manual intervention when agents are working through longer processes. That matters as companies move from simple AI chatbots towards systems that can carry out tasks across software applications.
Lower pricing and wider availability
Google is offering Gemini 3.7 Flash at an introductory price of $0.75 per 1 million input tokens and $3.75 per 1 million output tokens through the end of 2026. The lower pricing could make the model particularly attractive for applications that generate large volumes of AI requests.
Gemini Spark, Google's personal AI agent available to Google AI Pro and Ultra subscribers in more than 160 countries, is also moving to Gemini 3.7 Flash. Google says this will help Spark handle multi-step knowledge work across Google Workspace more efficiently.
Safety remains part of the rollout
Google says Gemini 3.7 Flash includes expanded protections against misuse in areas including chemical, biological, radiological and nuclear threats, as well as cyber offences. The company has also published a model card outlining the system's capabilities and limitations.
These safeguards are particularly relevant as increasingly capable models are used to perform tasks with greater autonomy. Gemini 3.7 Flash ultimately represents Google's push to make advanced AI more practical for everyday development and business workflows.
Its combination of stronger coding, improved reasoning, agent capabilities and lower pricing could make it a significant option for teams looking to scale AI applications without dramatically increasing costs.