Google just broke its own update cycle. If we were used to seeing half-point jumps in its artificial intelligence models, the company has decided to launch Gemini 3.1 Pro. This unexpected version does not come to add simple visual tricks, but rather has a clear objective: to confront complex problems where a simple answer simply is not enough.
According to the company, this model integrates the advanced reasoning engine that we saw debut recently, but bringing it to a much broader audience. In practice, this means that the tool is now capable of synthesizing large volumes of data into a single view, generating animations through code, or solve scientific challenges. However, it's not all magic: although Google boasts that it outperforms the competition in synthetic metrics, the real test will be in the daily lives of users, who often encounter hallucinations or practical limitations.
Beyond testing: the real impact for the user
The numbers on paper sound impressive. The model achieves a score of 77.1% in ARC-AGI-2, which means double the performance in abstract reasoning compared to the previous generation. There is one sector where it will shine: autonomous scheduling and independent agent workflows. Here, the AI must solve tasks in multiple steps without losing track, which is essential for developers.
At the moment, the new system is launched in the preview phase. Users can start testing it today in the main app and in NotebookLMalthough with more generous usage limits for paying subscribers and developers through its API.
It seems that the machinery does not rest: just yesterday Google announced the ability of Gemini to create songsand now presents Gemini 3.1 Pro that comes to give everything against its rivals. Will this be the model that finally convinces you to delegate your most critical tasks to a machine?
