Gemini 3 Pro and Nano Banana 2 models were seen; Is the supply close?

According to new evidence, Google will soon introduce two new models of artificial intelligence with the names Gemini 3 Peru and Nano Banana 2 will supply According to reports, these two models represent a new stage in the development of Google’s artificial intelligence technology; Where Gemini 3 is designed to process code and perform general tasks, Nano Banana 2 focuses on producing realistic and natural images.

Gemini 3 Peru and Nano Banana 2 Google is coming

Based on data published by the TestingCatalog account on X, the model Gemini-3-Pro Seen in Vertex AI, Google’s dedicated cloud infrastructure for building artificial intelligence products. The technical title of this model, “gemini-3-pro-preview-11-2025”, confirms that Google plans to release its new version in November 2025(November) start This is consistent with previous reports about the timing of the release.

Although some users on the X social network have claimed to have accessed Gemina 3, the technical evidence does not support this. Google usually conducts its internal testing in secret, involving only a limited group of business partners in the process. According to the non-disclosure agreements (NDA), these partners are not allowed to publish the results on any public platform.

New information shows that Gemina 3 will be released with a context limit of one million and is expected to be gradually available to users from November. Right now, Google’s latest AI model, Gemina 2.5 Pro, which has been out for 8 months, is still considered one of the most powerful AI models on the market. This model scored 63.8% in the SWE-Bench Verified benchmark, which is a global standard for measuring the performance of agent-based models, while its main competitor, the Claude Sonnet 4.5, managed to score around 77%.

At the same time, Google on a new model named Nano Banana 2 It works with the GEMPIX2 passcode, which was recently spotted on Gemina’s official website. This model is expected to be optimized to produce high-quality and more natural images, and according to the plan, in December 2025 be published

In the same period of time, OpenAI is also not sitting idle and is preparing GPT 5.1 and new versions of Codex; A platform known as the company’s agent-based coding terminal. The competition between Google and OpenAI has now reached a stage where both companies are trying to expand the boundaries of artificial intelligence capabilities and introduce a new generation of multipurpose and intelligent models to the market.