Google just launched its best AI model for coding
What's the story
Google has unveiled its latest artificial intelligence (AI) model, Gemini 2.5 Pro Preview (I/O edition). The updated version of the flagship Gemini 2.5 Pro is said to outperform some of the most popular benchmarks in the field. The new model is now available via Gemini API, Google's Vertex AI and AI Studio platforms, and Google's chatbot app for web and mobile devices. Demis Hassabis, CEO and co-founder of Google Deepmind, called it the company's "best coding model."
Model improvements
Enhanced coding and web app development capabilities
Google has claimed that Gemini 2.5 Pro Preview (I/O edition) offers significantly improved capabilities for coding and developing interactive web apps. The model also excels in tasks such as code transformation and editing. According to Google, this new version will not only enhance coding performance but also address major developer feedback including reducing errors in function calling and improving function calling trigger rates.
Benchmark performance
Gemini 2.5 Pro Preview leads in web app creation
In a blog post, Google revealed that the Gemini 2.5 Pro Preview (I/O edition) tops the WebDev Arena Leaderboard, a benchmark evaluating a model's capability to create visually appealing and functional web apps. The new model also shows outstanding performance in video understanding, scoring an impressive 84.8% on the VideoMME benchmark.
Developer conference
Google to showcase AI advancements at I/O developer conference
The launch of the Gemini 2.5 Pro Preview (I/O edition) comes just ahead of Google's annual I/O developer conference, where the tech giant is expected to unveil a slew of new models and AI-driven tools and platforms. The move comes as Google competes with other AI powerhouses like OpenAI, Anthropic, and xAI in an increasingly competitive market. Google I/O 2025 will be held on May 20-21, both in person at Shoreline Amphitheatre in Mountain View, California, and online via livestream.