From the archive
Gemini 2.5 Pro adds reasoning to a long-context model
Google releases an experimental Gemini 2.5 Pro model in AI Studio and Gemini Advanced, with Vertex AI planned to follow.
Google DeepMind introduced Gemini 2.5 Pro Experimental on March 25, 2025. The model added a reasoning stage before answering and began with a one-million-token context window.
Reasoning with mixed inputs
Gemini 2.5 Pro combined reasoning with support for text, audio, images, video, and code repositories. Google focused the announcement on harder problems and coding tasks, including generating applications and editing existing code. The longer context allowed substantial source material to accompany a request.
Experimental access first
The model became available in Google AI Studio and to Gemini Advanced subscribers in the Gemini app. Google Cloud Vertex AI availability and pricing for higher-volume use were scheduled to follow.
The release therefore opened an experimental route for testing the new model before its wider production rollout. Google also announced plans to increase the context window to two million tokens later.