Google Opens Free Access to Its Flagship Gemini 3 AI Model

Google has introduced the next generation of its language models—Gemini 3. The flagship version is already available for free testing to anyone through the web service AI Studio. The developers claim a significant leap in the model's ability for logical reasoning and solving complex problems. Google's internal tests show that Gemini 3 Pro achieves high scores in solving math problems, analyzing text, and especially in programming.

Some users with early access have already tested its coding capabilities. Examples have appeared online where the model generated working prototypes of famous games like Flappy Bird or the classic GTA from a single text prompt.

In specialized disciplines, the model also demonstrates leading performance:

You can try the model out via Google AI Studio; all you need is to sign in with a standard Google account. Alongside this, Google is rolling out a major update to its mobile Gemini app, where the new model is also gradually becoming available. For handling the most complex queries, the app features a special "Deep Think" mode.

With the release of Gemini 3, the AI industry continues to actively develop. The new Grok 4.1 model from xAI also recently became available to all users, marking the beginning of a new round of competition between the leading neural networks.