Mountain View, California — Google has unveiled the Gemini 3 Flash model, a rapid and cost-effective artificial intelligence system, as it seeks to bolster its competitive edge against rivals like OpenAI. This new model arrives just a month after the introduction of the Gemini 3 and aims to enhance user experience in its Gemini app and the AI functionality in Google Search.

The Gemini 3 Flash model, an upgrade over the previously released Gemini 2.5 Flash, exhibits impressive performance enhancements. It has outperformed earlier iterations and matched or surpassed other leading models in benchmark tests. For example, it achieved a score of 33.7% on the Humanity’s Last Exam benchmark, which evaluates expertise across various domains. By comparison, the Gemini 3 Pro and GPT 5.2 scored 37.5% and 34.5%, respectively, while Gemini 2.5 Flash only garnered 11%.

In addition to its strong performance in standard tests, the new model demonstrated exceptional capabilities in multimodal tasks, scoring 81.2% on the MMMU-Pro benchmark—significantly outpacing its competitors. This positions the model as a leader in its category, making it particularly appealing for diverse applications.

Google has made the Gemini 3 Flash the default offering in its Gemini app, replacing Gemini 2.5 Flash worldwide. Users still have the option to select the Pro model for specific tasks, such as advanced math and coding queries. The company highlights the new model’s aptitude for processing multimodal information, allowing users to upload various media types—ranging from videos to audio recordings— and receive contextually relevant responses.

The Gemini 3 Flash is also designed to better interpret user queries, leading to more visually engaging answers that incorporate elements like images and tables. Furthermore, it enables users to prototype applications directly within the app using simple prompts, expanding its functionality beyond traditional AI capabilities.

For enterprises, Google has revealed that well-known companies like JetBrains and Figma have begun incorporating the Gemini 3 Flash model into their operations, accessible via Vertex AI and Gemini Enterprise. Developers can also access the model in preview mode through API functionalities, which are integrated into Google’s recently launched coding tool, Antigravity.

In performance benchmarks for coding tasks, the Gemini 3 Pro achieved a notable 78% on the SWE-bench, ranking just behind GPT-5.2. Google suggests that this model is optimized for a variety of uses including video analysis, data extraction, and fast-paced workflows. Despite slightly higher costs compared to previous models, Google asserts that the new system delivers greater efficiency, being significantly faster and utilizing fewer tokens on average.

The current pricing structure stands at $0.50 per 1 million input tokens and $3.00 per 1 million output tokens, a slight increase from previous rates. Google’s leaders argue that even with the modest rise in costs, businesses will ultimately benefit from the system’s enhanced performance and speed.

As part of its ongoing competition in the AI landscape, Google has reported processing over 1 trillion tokens daily through its API, signaling robust demand for its AI offerings. This competitive environment is further intensified by actions from rivals like OpenAI, which has recently launched its own models following a reported internal alert regarding market dynamics.

While Google has refrained from making direct comparisons with OpenAI, it acknowledges the influence of these developments on the industry. Executives emphasize that continuous advancements in AI not only push their own innovations but also set new benchmarks for evaluating performance across the sector.