In an unexpected move, Google has officially launched the Gemini 2.5 Pro Preview for developers, preceding the anticipated unveiling at the upcoming Google I/O 2025. This early release was driven by substantial demand from the developer community, eager to harness the capabilities of this advanced AI model.
The Gemini 2.5 Pro Preview builds upon the already robust framework of its predecessor, the original Gemini 2.5 Pro. Notably, it has achieved a remarkable performance increase, surpassing the earlier version by an impressive 147 Elo points on the WebDev Arena leaderboard.
Google highlights that Gemini 2.5 Pro possesses a “deep understanding of code” and exhibits exceptional reasoning skills, making it the model of choice for developers. In addition to its coding prowess, the model demonstrates significant advancements in video comprehension, attaining a score of 84.8 per cent on the VideoMME benchmark.
This cutting-edge model is the driving force behind innovative tools such as Cursor’s code agent. Google has also forged partnerships with leading companies like Cognition and Replit to explore and expand the horizons of developer capabilities.
Developers can seamlessly access the enhanced features of the Gemini 2.5 Pro through platforms like Google AI Studio or Vertex AI. Users of the previous version are automatically redirected to the latest iteration, ensuring a smooth transition without any additional costs or procedures.
Yorum Yap