Google introduced Gemini 2.5, its latest experimental artificial intelligence model

it ranked second in a competitive leaderboard for AI-driven web development tools.

Google's latest experimental artificial intelligence model, Gemini 2.5, ranked second in a competitive leaderboard for AI-driven web development tools, according to the AI benchmarking platform LMArena.

The new AI model topped the charts in reasoning and knowledge, science and mathematics, rendering it Google's "most intelligent AI model," the company claimed.

On March 25, Google announced that it will allow developers to try out Gemini 2.5 Pro, which the company described as a thinking model, capable of considering thoughts before responding.

This improves both accuracy and performance, particularly in coding, science and math tasks, and enables Gemini 2.5 to support more context-aware agents, it said.

Google's new model ranks second in web dev leaderboard

The new AI model had an arena score of 1267.70, surpassing competitors including DeepSeek, Grok and ChatGPT. However, the top spot went to Anthropic's AI model Claude 3.7 Sonnet, which had an arena score of 1354.01.

The models were ranked in real-time in LMArena's WebDev leaderboard, a real-time AI coding competition where models compete in web development challenges created by the AI benchmarking platform.