Google appears to be preparing for the introduction of Gemini 3. The new AI model briefly appeared in AI Studio, the development platform where programmers, researchers, and students work with the various Gemini models.
According to BleepingComputer, the fact that the model is now visible there indicates that the official rollout is not far off. The update is expected to go live within a few hours to days.
AI Studio is an important place for Google to introduce new models in a controlled manner. Users can experiment with settings such as context length and temperature, making it a logical first environment for the phased release of Gemini 3. At present, Gemini 2.5 Pro is still considered the most advanced model within the tool, but the interface already shows subtle adjustments that seem to be aimed at the upcoming generation.
Versatile and efficient model
With Gemini 3, Google also wants to further intensify competition with other AI providers. The market is currently dominated by parties such as OpenAI and Anthropic, and Google sees the new version of Gemini as an opportunity to regain ground. According to CEO Sundar Pichai, Gemini 3 will be a significantly more powerful AI agent than previous models. He emphasizes that the leap forward is greater than what has been achieved in recent years and that Google is combining multiple technologies to make the model more versatile and efficient.
One of the most striking clues in AI Studio concerns the temperature setting, a parameter that influences the creativity and consistency of model output. A new line of text suggests that Gemini 3 functions optimally at a temperature of 1.0. Lower values could negatively affect reasoning ability, indicating that Google has carefully tuned the model’s character for complex tasks.
In addition to AI Studio, references to the model are also popping up elsewhere. On the Vertex AI cloud platform, which is used to develop AI agents and related applications, a variant was previously spotted under the name Gemini-3-pro-preview-11-2025. This indicates that Google is already conducting advanced internal testing and that the company is preparing different configurations for developers.
Gemini 3 is not the only newcomer. Google is simultaneously working on Nano Banana 2, a model that, according to notable mentions on the Gemini website, appears to be focused on advanced image generation. Internally, the project is codenamed GEMPIX2. If the signals are correct, this image generator could be released as early as December 2025.
The early visibility of Gemini 3 in AI Studio suggests that its launch is very close. Google has indicated that it will soon share more details about the new generation of models, suggesting that the coming days will be crucial for the further rollout strategy.