Google Unveils Gemini 2.5: The Most Advanced Reasoning AI Yet

Google announced the official release of Gemini 2.5, a new reasoning model that the company claims is its most intelligent AI model to date.

DeepMind CTO Koray Kavukcuoglu described Gemini 2.5 as a "thinking model" that performs reasoning before responding, improving performance and accuracy. In AI, reasoning means analyzing information, drawing logical conclusions, incorporating context and nuances, and making informed decisions.

Gemini 2.0 Flash Thinking was Google’s first reasoning model; Gemini 2.5 builds on it with an improved base model and fine‑tuning. Google says future AI models will all have built‑in reasoning capabilities.

The first Gemini 2.5 model, Gemini 2.5 Pro Experimental, far outperforms competitors such as OpenAI o3‑mini, Claude 3.5 Sonnet, and DeepSeek R1 on the LMArena benchmark.

It achieved an 18.8% score on the Humanity’s Last Exam dataset—a test designed by hundreds of subject‑matter experts to capture the frontier of human knowledge and reasoning. The model also excels at coding, especially creating web and agent applications and handling code translation. By comparison, OpenAI o3‑mini scored 14% and DeepSeek R1 scored 8.6%.

The model is now available to premium subscribers in Google AI Studio and the Gemini app. Google is working to add it to Vertex AI and will announce pricing in the coming weeks.

At launch it offers a 1 million‑token context window, and Google is striving to expand this to a 2 million‑token window as soon as possible.