Google has launched its rapid and economical Gemini 3 Flash model today, building on the Gemini 3 released last month, in a move to rival OpenAI. This new model will also become the default in the Gemini app and for AI mode within search.
This Flash model arrives six months after Google introduced the Gemini 2.5 Flash model, bringing substantial enhancements. In benchmark tests, the Gemini 3 Flash model demonstrates significant improvements over its predecessor and, in some metrics, achieves performance comparable to other leading models like Gemini 3 Pro and GPT 5.2.
For example, it achieved a 33.7% score without tool usage on Humanity’s Last Exam benchmark, which assesses expertise across various fields. In contrast, Gemini 3 Pro scored 37.5%, Gemini 2.5 Flash scored 11%, and the recently launched GPT-5.2 scored 34.5%.
On the MMMU-Pro multimodality and reasoning benchmark, the new model surpassed all competitors with an 81.2% score.
Consumer rollout
Google is establishing Gemini 3 Flash as the primary model in the Gemini app globally, replacing Gemini 2.5 Flash. Users retain the option to select the Pro model via the model picker for advanced math and coding inquiries.
The company states that the new model is adept at recognizing multimodal content and generating relevant answers. For instance, users can upload a pickleball video to receive tips, sketch a drawing for the model to identify, or upload an audio recording for analysis or quiz generation.
The company also mentioned that the model has an enhanced understanding of user query intent and can produce more visual responses, including elements like images and tables.
You can also leverage the new model to develop app prototypes within the Gemini app using prompts.
Gemini 3 Pro is now accessible to all users in the U.S. for search, and a broader audience in the U.S. can now utilize the Nano Banana Pro image model in search as well.
Enterprise and developer availability
Google noted that companies such as JetBrains, Figma, Cursor, Harvey, and Latitude are already employing the Gemini 3 Flash model, which is accessible through Vertex AI and Gemini Enterprise.
For developers, the model is available as a preview through the API and within Antigravity, Google’s new coding tool introduced last month.
The company stated that Gemini 3 Pro achieved a 78% score on the SWE-bench verified coding benchmark, with only GPT-5.2 outperforming it. It further added that the model is well-suited for video analysis, data extraction, and visual Q&A, and its high speed makes it ideal for quick and repetitive tasks.
Model pricing is set at $0.50 per 1 million input tokens and $3.00 per 1 million output tokens. This represents a slight increase from Gemini Flash 2.5’s rates of $0.30 per 1 million input tokens and $2.50 per 1 million output tokens. However, Google asserts that the new model offers superior performance to the Gemini 2.5 Pro model while operating three times faster. Moreover, for tasks requiring