Google released Gemini 3.7 Flash on August 13. The company calls it its most intelligent workhorse model yet for coding and agent tasks. The launch comes just three weeks after Gemini 3.6 Flash.
Google says the fast turnaround came from developer feedback and algorithmic improvements. The model card describes 3.7 Flash as a refinement of 3.6 Flash. There was no new pretraining run behind it.
The model accepts text, images, audio and video. It supports a 1 million token context window. It can return up to 64,000 output tokens. Developers can also adjust thinking settings to trade quality against cost and speed.
Pricing is a headline feature here. Introductory rates are 75 cents per million input tokens and 3.75 dollars per million output tokens. That is half the initial price of Gemini 3.6 Flash. Pricing rises on January 1, 2027, to 1.50 dollars per million input tokens and 7.50 dollars per million output tokens.
Gemini 3.7 Flash is rolling out across Google’s developer tools. It is now available in Gemini Spark for AI Pro and Ultra subscribers. It also reaches Google Antigravity, AI Studio, Android Studio, the Gemini Enterprise Agent Platform and the Gemini Enterprise app.
The release lands as competition in fast, cheap coding models keeps intensifying. Google, OpenAI and Chinese labs including Zhipu AI have all shipped new coding-focused models within days of each other this month. For developers, the price cut matters as much as the capability gains, since coding agents can burn through tokens quickly during long sessions.
Google’s launch sits alongside two other major releases this week. Industry coverage grouped Gemini 3.7 Flash with OpenAI’s Ultrafast and DeepSeek’s V4-Pro as three significant model updates arriving within days of each other. OpenAI’s Ultrafast reportedly delivers a 5.6-times speed gain on economically relevant knowledge work, though it remains limited to select customers in preview. DeepSeek’s V4-Pro, by comparison, prices its tokens higher than Gemini 3.7 Flash. But it ships as open-weight, giving developers a cheaper, inspectable alternative to Google’s closed model.
