Google has released what it calls its most intelligent AI model yet, Gemini 2.5. The first 2.5 model release, an experimental version of Gemini 2.5 Pro, is a next-gen reasoning model that Google says outperformed OpenAI o3-mini and Claude 3.7 Sonnet from Anthropic on common benchmarks “by meaningful margins.” Gemini 2.5 models “are thinking models, capable of reasoning through their thoughts before responding, resulting in enhanced performance and improved accuracy,” according to Google. The new model comes just three months after Google released Gemini 2.0 with reasoning and agentic capabilities.

Gemini 2.5 Pro is available now in Google AI Studio and in the Gemini app for those who subscribe to the $20 per month Gemini Advanced. Soon it will also be added to Google’s Vertex AI cloud-based AI developer platform.

“Like Gemini 2.0 and Gemini 2.0 Flash Thinking, Gemini 2.5 Pro ‘thinks’ before it responds,” writes VentureBeat, detailing that “the new model can handle multimodal input from text, audio, images, videos and large datasets” and can understand “entire code repositories for coding projects.”

“With Gemini 2.5, we’ve achieved a new level of performance by combining a significantly enhanced base model with improved post-training,” Google DeepMind CTO Koray Kavukcuoglu wrote in a blog post. “Going forward, we’re building these thinking capabilities directly into all of our models, so they can handle more complex problems and support even more capable, context-aware agents.”

VentureBeat points out that Gemini 2.5 Pro “offers some of the largest context windows available for experimental models on Gemini,” shipping with a 1 million token context window that Google plans to expand soon to 2 million tokens.

“Reasoning techniques have helped AI models achieve new heights in math and coding tasks,” writes TechCrunch. “Many in the tech world believe reasoning models will be a key component of AI agents,” which can autonomously perform tasks without human intervention.

Google has indicated all of its future AI models will include reasoning capabilities, first introduced with OpenAI o1 in September 2024. Since then, competing AI companies have rushed to match or surpass its capabilities. Today, in addition to OpenAI and Google, companies such as Anthropic, DeepSeek and xAI have their own reasoning models, “which use extra computing power and time,” making them more expensive to use, TechCrunch points out.

Google says it will unveil pricing in the coming weeks,” enabling people to use 2.5 Pro with higher rate limits for scaled production use.”