Google unveils Gemini 3.7 Flash AI model for coding, agent workflows

Alphabet's Google unveiled Gemini 3.7 Flash on Thursday, an AI model aimed at software coding and automated business tasks. However, the release of the company's premium Gemini 3.5 Pro model remains uncertain. Investors have been monitoring Gemini 3.5 Pro's progress as a benchmark for DeepMind's ability to compete with Anthropic and OpenAI. In July, Google had stated that Gemini 3.5 Pro was being piloted with partners and would be launched "soon."

Gemini 3.7 Flash is positioned as a cost-effective solution for businesses developing autonomous AI systems capable of planning tasks, utilizing software tools, and executing multi-step workflows with minimal human involvement. The model was released three weeks after Gemini 3.6 Flash. Google reported improved performance on coding tasks, such as debugging, issue resolution, and generating production-ready code, in a company blog post.

To encourage adoption, Google is offering Gemini 3.7 Flash at an introductory price of $0.00075 per million input tokens and $0.00375 per million output tokens through the end of the year, which is half the original cost of Gemini 3.6 Flash. The model is now available immediately to Gemini Spark, Google's subscription-based AI agent service accessible to Google AI Pro and Ultra customers in over 160 countries.

Google co-founder Sergey Brin has been encouraging key AI personnel to fully commit to the Gemini model, as Alphabet strives to narrow the gap with rivals, according to a report by Reuters on Wednesday. Last week, the company announced a significant leadership shake-up in its Google DeepMind AI division, with former chief Demis Hassabis stepping down and being replaced by his deputy, Koray Kavukcuoglu.

Two original technical co-leads of Gemini also departed to found a startup. CEO Sundar Pichai defended Google's AI strategy during the company's July earnings call, addressing concerns about falling behind rivals due to delays in releasing the flagship model and losing ground in AI coding.