Google launches Gemini 3 Flash, makes it the default model in the Gemini app

Google has officially launched its latest AI model, Gemini 3 Flash, aiming to dominate the competitive landscape previously led by OpenAI. This new iteration, which is now the default option in the Gemini app and integrated into AI search functionalities, is positioned as a faster and more cost-effective solution compared to its predecessors. Arriving just six months after the Gemini 2.5 Flash, the Gemini 3 Flash boasts significant advancements in performance. In benchmark tests, it outshines earlier models, even rivaling other leading systems like Gemini 3 Pro and GPT 5.2 in specific assessments. For example, it achieved a score of 33.7% on the Humanity’s Last Exam benchmark, which evaluates expertise across various domains. In comparison, the Gemini 3 Pro scored 37.5%, while GPT-5.2 reached 34.5%, and Gemini 2.5 Flash lagged behind at 11%. Additionally, the Gemini 3 Flash excelled in the multimodality and reasoning benchmark MMMU-Pro, securing an impressive score of 81.2%. This model is now the preferred choice for users globally, replacing the older Gemini 2.5 Flash, although users still have the option to select the Pro model for specialized tasks like math and coding. One standout feature of the Gemini 3 Flash is its enhanced capability to interpret multimodal content. Users can upload a variety of formats, such as videos or audio recordings, and receive tailored analyses or advice. For instance, if someone shares a short video of their pickleball game, the model can offer tips, or it can identify sketches and generate quizzes based on uploaded audio. The new model also promises a better understanding of user queries, allowing it to produce rich visual responses that include images and tables. Furthermore, it facilitates app prototype creation directly within the Gemini app, enhancing user experience. In the U.S., the Gemini 3 Pro is now accessible to a wider audience, and the advanced Nano Banana Pro image model is also available. Notably, several companies, including JetBrains and Figma, have already begun utilizing the Gemini 3 Flash model through Vertex AI and Gemini Enterprise. Developers can also access the model in a preview through the API and Google’s new coding tool, Antigravity. The Gemini 3 Pro has scored 78% on the SWE-bench verified coding benchmark, second only to GPT-5.2. It is particularly effective for tasks such as video analysis and visual Q&A, with its rapid processing speed making it ideal for repetitive workflows. Pricing for the new model is set at $0.50 per million input tokens and $3.00 per million output tokens, which is slightly higher than the previous version. Despite the increased cost, Google asserts that the Gemini 3 Flash outperforms the Gemini 2.5 Pro model while being three times faster and utilizing 30% fewer tokens on average for cognitive tasks. Tulsee Doshi, Senior Director & Head of Product for Gemini Models, described the Flash model as a more economical option for businesses needing to handle bulk tasks efficiently. Since the launch of Gemini 3, Google's API has been processing over 1 trillion tokens daily, intensifying the competition with OpenAI. Following a notable decline in ChatGPT's traffic, OpenAI responded with the release of GPT-5.2 and a new image generation model. While Google did not directly comment on its rivalry with OpenAI, it acknowledged the dynamic nature of the industry and the ongoing improvements being made by all players involved.