Google continues its investments in artificial intelligence at full speed. According to sources close to the company, the tech giant is preparing to launch two new AI models, the Gemini 3 Pro and the Nano Banana 2, soon.
According to leaked information, the Gemini 3 Pro will focus on coding and general use, while the Nano Banana 2 is being developed to produce high-quality, lifelike images.
Gemini 3 Pro May Be Unveiled in November
According to data revealed on the X platform, the model, dubbed gemini-3-pro-preview-11-2025, has been captured for the first time on Google’s Vertex AI system. This detail largely confirms previous leaks that the Gemini 3 Pro will be announced in November.
While some users claim to have gained early access to the new model, Google is generally known to proceed with closed testing. The company likely granted limited access to select partners, but it’s notable that these partners are unable to publicly disclose their results due to confidentiality agreements.
According to leaked technical information, the Gemini 3 Pro will come with support for 1 million context windows. This will allow the model to analyze much longer code snippets, documents, or conversations at once.
The current version, Gemini 2.5 Pro, was released approximately eight months ago and remains one of the most powerful models in the industry. The model achieved a 63.8% success rate in the SWE-Bench Verified test with a custom agent setup. In the same test, the Claude Sonnet 4.5 model topped the list with a 77% success rate.
Google’s second new model is poised to revolutionize image production. The Nano Banana 2, codenamed GEMPIX2, was recently spotted on Gemini’s official website.
This development suggests that the image production model could be available in December 2025.
The new model is reportedly optimized for creating high-resolution, lifelike images. It is also expected to integrate with Google’s existing generative AI infrastructure, serving as a powerful visual engine for mobile and web applications.
Google’s announcement of two new AI models before the end of the year signals a renewed competition with OpenAI. OpenAI is also working intensively on GPT 5.1 and the new Codex models. According to industry experts, a major leap forward will be made among AI models in the coming months, both in terms of coding capabilities and visual production power.
