Today, Google introduced "Gemini 3 Flash," a new, fast, and cost-effective artificial intelligence model. This new version, built on the Gemini 3 platform released last month, is now available globally in the Gemini app and as the default model in the "AI mode" section of Google Search.
Performance Improvements
The new model is significantly more powerful than its predecessor, "Gemini 2.5 Flash." According to test results, it can even compete with "Gemini 3 Pro" and OpenAI’s latest "GPT-5.2" model in certain metrics. For example, in the "MMMU-Pro" test, which requires complex reasoning and analysis, it achieved a score of 81.2%, outperforming all its competitors.
User Capabilities
The company states that "Gemini 3 Flash" understands user intent more accurately and can analyze multimodal content (video, audio, images) more effectively. For instance, users can upload a video and receive advice, ask about a simple hand-drawn sketch, or analyze a voice recording. Additionally, the new model allows users to create simple mobile app prototypes.
For Business and Developers
Companies such as JetBrains, Figma, and Cursor have already started using this model. In terms of pricing, it is slightly more expensive than the previous Flash version (0.50 USD per 1 million input tokens), but Google claims it is more powerful than "Gemini 2.5 Pro" and three times faster. Because it uses fewer tokens for tasks requiring reasoning, it is considered more cost-efficient overall.
This move continues Google’s fierce competition with OpenAI. It’s worth noting that recently OpenAI issued an internal “alarm” due to decreased traffic and promptly released the GPT-5.2 model.
