Google has introduced Gemini 2.5 Pro Preview (I/O edition), a new version of its flagship Gemini 2.5 Pro AI model. The update arrives before the company’s annual I/O developer conference and is aimed squarely at developers building with AI.
The company says the model improves on coding, web app creation, code transformation, code editing, function calling, and video understanding. It is available through the Gemini API, Vertex AI, AI Studio, and the Gemini chatbot app on web and mobile devices.
A developer-focused Gemini update
Gemini 2.5 Pro Preview (I/O edition) effectively replaces Gemini 2.5 Pro while keeping the same price. That matters for teams already using the model because the update is positioned as a performance improvement rather than a new premium tier.
Google’s emphasis is clear: this version is built to be stronger for software work. The company says the model has significantly improved capabilities for coding and for building interactive web apps. It also says the model performs better at code transformation, meaning it can modify existing code to reach a specific goal.
Code editing is another stated area of improvement. For developers, that points to a model designed not only to generate fresh code, but also to revise, adapt, and refine code that already exists.
Where the model is available
Google is making Gemini 2.5 Pro Preview (I/O edition) available across several of its AI products and developer platforms. That includes the Gemini API, Google’s Vertex AI platform, AI Studio, and the Gemini chatbot app for both web and mobile devices.
This broad rollout means the model is not limited to one type of user interface. Developers can access it through API-based workflows, enterprise-oriented cloud tools, experimentation environments, or the consumer-facing Gemini app.
The pricing is the same as Gemini 2.5 Pro, the model it effectively replaces. The source article does not give specific prices, but the unchanged pricing point is important because Google is framing the update as a direct successor rather than a separate, differently priced product.
Benchmarks and claimed strengths
Google claims Gemini 2.5 Pro Preview (I/O edition) tops a number of widely used benchmarks. In a blog post, Google notes that the model leads the WebDev Arena Leaderboard, which measures a model’s ability to create web apps that are both aesthetically pleasing and functional.
The model is also described as having state-of-the-art performance in video understanding. Google says it achieved a score of 84.8% on VideoMME, a popular benchmark for that category.
Demis Hassabis also highlighted the launch on May 6, 2025, saying the model was Google’s best coding model so far and noting that it ranks no.1 on LMArena in Coding and no.1 on the WebDev Arena Leaderboard. Those claims reinforce the same central message: Google wants this Gemini release to be judged heavily on developer tasks.
Function calling and web development
Beyond benchmark results, Google says the update addresses feedback from developers already using Gemini 2.5 Pro. The company specifically points to reducing errors in function calling and improving function calling trigger rates.
Function calling matters because it is one of the ways AI models interact with tools, services, and structured software workflows. If a model calls functions incorrectly, or fails to call them when it should, developers can face unreliable behavior in applications that depend on AI-driven actions.
Google also describes the model as having a strong preference for aesthetic web development while keeping its steerability. In practical terms, the company is presenting Gemini 2.5 Pro Preview (I/O edition) as a model that can help produce more polished web app outputs without taking control away from the developer’s instructions.
Why the timing matters
The release comes ahead of Google’s annual I/O developer conference, which explains the I/O edition name. The source article says Google is expected to unveil a host of models, along with AI-powered tools and platforms, at the event.
The launch also lands during an intense AI race. Google is competing for both mindshare and marketshare, while rivals such as OpenAI and xAI are described as being on the cusp of releasing models expected to be highly performant.
For Google, this Gemini update gives developers something concrete before I/O: a refreshed flagship model with stronger coding claims, broader web app capabilities, the same pricing as its predecessor, and availability across the company’s main AI access points.