Gemini 2.5 Pro and Flash stable, new Flash-Lite

Google has announced two major new features for the Gemini 2.5 family of models. After a few months of testing and improvements, the Mountain View company has released the stable versions of Gemini 2.5 Pro and Flash . At the same time, the new Gemini 2.5 Flash-Lite model has arrived (preview). Users can alsoupload videos to the Gemini app for Android and iOS.

What's new for the Gemini 2.5 family

Gemini 2.5 Pro was announced on March 25, while Gemini 2.5 Flash has been available since April 17. Initially, these were “preview” (experimental) versions, so they did not guarantee 100% reliable results. After making several improvements, also based on feedback received from developers, Google released the stable versions that can be selected in the Gemini app.

In detail, the stable version of Gemini 2.5 Pro is based on the June 5 preview, while the Gemini 2.5 Flash is based on the May 20 preview. Both are models with reasoning capabilities, but they differ in size, performance and price. For Gemini 2.5 Flash, it is possible to disable the reasoning capability.

The latest addition to the family is the Gemini 2.5 Flash-Lite (preview). It offers better performance and responsiveness than the previous Gemini 2.0 Flash-Lite in several tasks (math, programming, science, multimodality). It has a context window of one million tokens and can be connected to Google Search.

Developers can use specific APIs to assign a reasoning budget. Unlike other models, reasoning is turned off by default to provide higher performance at lower cost. The Gemini 2.5 Flash-Lite preview is available through Google AI Studio and Vertex AI. Custom versions of Gemini 2.5 Flash and 2.5 Flash-Lite are used in Google Search for AI Overview and AI Mode.

The Gemini app now supports video uploads (currently only on Android and iOS), as well as documents and images. After uploading the video, you can ask the chatbot for a description.