Google Unveils Gemini 2.5 Pro I/O Edition with Major Boost to Coding Capabilities

Just days before the highly anticipated Google I/O 2025 event, Google has unveiled a significant update to its flagship AI model: Gemini 2.5 Pro (I/O Edition). Billed as a “pre-I/O goodie” by Google DeepMind CEO Demis Hassabis, the upgraded model brings substantial advancements in programming and software development capabilities. “Massively improved coding capabilities,” said Hassabis, announcing the release on May 6, two weeks ahead of schedule.

Initially slated for announcement during I/O 2025 on May 20, the Gemini 2.5 Pro (Preview)—labelled gemini-2.5-pro-preview-05-06—has now gone live across Google’s Gemini platforms. Users can access the model through the Gemini app (notably enhancing the Canvas workspace) and via the Gemini API, Google AI Studio, and Vertex AI.

What’s New in Gemini 2.5 Pro I/O Edition?

The model builds on the foundation of the first Gemini 2.5 Pro, released in March, with notable improvements in code transformation, editing, and the development of complex agentic workflows. Google emphasises that the tool now performs significantly better in real-world coding scenarios, particularly in building interactive web apps and agent-driven systems.

Google describes the release as “a big leap over 2.0” and notes that “more improvements [are] to come.” The Gemini 2.5 Pro I/O Edition scored 63.8% on SWE-Bench Verified, a benchmark evaluating autonomous software engineering. Using an agent-based methodology, the model demonstrated increasing strength in handling nuanced programming tasks with minimal human intervention.

A Model Built for Thoughtful AI Performance

Each model in the Gemini 2.5 lineup is designed as a “thinking model”—capable of reasoning through problems before responding. Google claims this results in better accuracy and overall AI performance. The I/O Edition reportedly tops the LMArena leaderboard by a wide margin, reflecting superior user preference, and ranks first in academic benchmarks such as AIME 2025 (math) and GPQA diamond (science).

What sets Gemini 2.5 Pro apart is its ability to deliver these high scores without test-time tricks like majority voting, which often increase processing costs. With expectations running high for new announcements at I/O 2025—including Android 16, Android XR, and broader Gemini integrations—this early release suggests Google is ready to lead the next wave in AI-assisted coding.