Google has officially launched its new Gemini 3 Flash model, a fast and cost-effective AI solution based on the previously released Gemini 3. This strategic move aims to intensify competition with OpenAI. The Gemini 3 Flash is now set as the default model within the Gemini app and for the AI mode in Google Search, marking a significant upgrade from the Gemini 2.5 Flash model announced six months prior.

On critical benchmarks, the Gemini 3 Flash model demonstrates substantial improvements, surpassing its predecessor and matching the performance of other frontier models like Gemini 3 Pro and GPT 5.2 in specific areas. For instance, on Humanity’s Last Exam benchmark, designed to evaluate expertise across diverse domains, Gemini 3 Flash scored 33.7% without tool use. This compares favorably to Gemini 3 Pro's 37.5%, GPT-5.2's 34.5%, and significantly outperforms Gemini 2.5 Flash's 11%. Furthermore, on the MMMU-Pro multimodality and reasoning benchmark, the new model achieved an impressive 81.2% score, outperforming all its competitors.

For consumers, the Gemini 3 Flash is now the global default in the Gemini app, replacing Gemini 2.5 Flash, though users can still select the Pro model for complex math and coding queries. Google highlights the model's enhanced ability to identify multimodal content and generate relevant answers. Practical applications include analyzing pickleball videos for tips, guessing objects from user sketches, or providing analysis and quizzes from audio recordings. The model is also designed to better understand user intent and produce more visual answers, incorporating elements like images and tables. Additionally, users can leverage the Gemini app to create app prototypes using prompts. Beyond Flash, the Gemini 3 Pro is now widely available in the U.S. for search, and more U.S. users can access the Nano Banana Pro image model in search.

Enterprise and developer communities are already adopting Gemini 3 Flash, with companies such as JetBrains, Figma, Cursor, Harvey, and Latitude utilizing the model. It is accessible through Vertex AI and Gemini Enterprise. For developers, a preview of the model is available via API and within Antigravity, Google’s new coding tool. Google notes that Gemini 3 Pro scored 78% on the SWE-bench verified coding benchmark, a performance only surpassed by GPT-5.2. The model is particularly well-suited for video analysis, data extraction, and visual Q&A, and its speed makes it ideal for rapid and repeatable workflows.

The pricing for Gemini 3 Flash is set at $0.50 per 1 million input tokens and $3.00 per 1 million output tokens. This is a slight increase from Gemini 2.5 Flash's pricing of $0.30 per 1 million input tokens and $2.50 per 1 million output tokens. However, Google asserts that the new model outperforms Gemini 2.5 Pro while being three times faster. For