Google has officially announced that they have released the Gemini 2.5 Pro Preview for developers. Originally scheduled to launch at the Google I/O event on May 20 and 21, the advanced AI model was launched early due to an "overwhelming response" from the developer community.
The new Gemini 2.5 Pro Preview version is based on the strong feedback for the original Gemini 2.5 Pro. In the WebDev Arena leaderboard—a benchmark based on human preferences on how well an AI model can create attractive and functional web apps—the updated version outperforms its predecessor by 147 Elo points.
Google states that Gemini 2.5 Pro has a "deep understanding of code" plus powerful reasoning capabilities, making it the go-to model for developers. The new model also excels in video understanding and has scored 84.8% on the VideoMME benchmark.
It powers innovative tools like Cursor"s code agent and is used in partnership with companies like Cognition and Replit to explore new possibilities for developers. Silas Alberti from Cognition said that "It felt like a more senior developer because it was able to make correct judgement calls and choose good abstractions."
Replit"s Michele Catasta said that they found "Gemini 2.5 Pro to be the best frontier model when it comes to "capability over latency" ratio." In the official blog, Google has also highlighted some of the model"s use cases, such as turning YouTube videos into full learning apps, generating styled UI components, and creating web apps from simple ideas.
Developers can access the new version of the Gemini 2.5 Pro model through Google AI Studio or Vertex AI. Accessing the previous version automatically redirects to the latest release, with no extra cost or steps required.