Google has released 'Gemini 3.7 Flash,' offering performance comparable to Claude Sonnet 5 and GPT-5.6 Terra, and will be available at half the price of Gemini 3.6 Flash until the end of the year.



On August 13, 2026, local time, Google released Gemini 3.7 Flash , its most intelligent AI model to date, developed for coding and agents.

Gemini 3.7 Flash: our most intelligent workhorse model

https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/

Just three weeks after releasing Gemini 3.6 Flash , Google has released Gemini 3.7 Flash. Gemini 3.7 Flash incorporates developer feedback and algorithmic innovations from Gemini 3.6 Flash. Gemini 3.7 Flash delivers significant improvements across the entire workflow of software engineering, knowledge work, and web development, and is available at half the price of Gemini 3.6 Flash.

The graph below compares Gemini 3.7 Flash with other AI models in FrontierCode, the most demanding benchmark test for evaluating AI coding agents. Gemini 3.7 Flash performs better than not only its predecessor, Gemini 3.6 Flash, but also than Anthropic's Claude Sonnet 5 and OpenAI's GPT-5.6 Terra.



The graph below compares the scores of DeepSWE, a software engineering benchmark test used to measure the performance of AI coding agents. While the score is lower than that of the GPT-5.6 Terra, the score of the Claude Sonnet 5 is higher, representing a significant performance improvement over its predecessor, the Gemini 3.6 Flash.



The following is a comparison of the scores from Code Arena, a blind test (human evaluation) used to measure the web development and coding capabilities of AI models. Gemini 3.7 Flash achieved the highest score among the models compared.



The following graph compares the scores of GDP.pdf, a benchmark test that measures the advanced document processing capabilities and multimodal reasoning abilities of AI.



The following is a comparison of the scores from AutomationBench, a practical benchmark test that measures how accurately AI agents can complete actual business workflows.



In web development, Gemini Flash 3.7 enables the creation of more functional layouts and feature-rich apps with fewer prompts. UI generation achieves high design compliance and consistency by relying on reference inputs such as screenshots, images, and design systems.

Gemini Flash 3.7 will be available until the end of 2026 at a special price of $0.75 (approximately 120 yen) per million input tokens and $3.75 (approximately 600 yen) per million output tokens.

Detailed benchmark results for Gemini Flash 3.7 are summarized below.



Google CEO Sundar Pichai stated, 'Our Flash model is a reliable workhorse that delivers great performance at an affordable price. We release updates quickly to get it into the hands of developers as soon as possible. And now, just three weeks after the release of Gemini 3.6 Flash, Gemini 3.7 Flash shows significant improvements, including in coding and agent work. It brings substantial improvements to software engineering, knowledge work and web development, all at an introductory price that is half the original cost of Gemini 3.6 Flash.'




Artificial Analysis, an independent AI benchmark site, has reported that Gemini 3.7 Flash scored '56' on its Artificial Analysis Intelligence Index, a composite benchmark test that measures the overall intelligence of AI models. This score is third only to GPT-5.6 Terra (57) and Muse Spark 1.2 (57).




Logan Kilpatrick, a software engineer and product lead for Google AI Studio, emphasized that Gemini 3.7 Flash is faster, 50% cheaper than Gemini 3.6 Flash (until the end of the year), and offers a significant performance improvement in just three weeks. He also highlighted that Gemini 3.7 Flash is available for use with APIs, AI Studio, and Antigravity.




Kolay Kavuktsiohl, Senior Vice President and Chief AI Architect at Google DeepMind, highlights the rapid growth of Gemini's Flash series, which has evolved from version 3.5 to 3.7 in just three months.




Related Posts:

in AI, Posted by logu_ii