What you need to know
- Google is releasing Gemini 2.5 starting today for Gemini Advanced subscribers.
- Gemini 2.5 Pro Experimental is the first model to ship, and it is an industry-leader in benchmarks.
- All Gemini 2.5 models are thinking models with native multimodality.
Google’s latest AI model, Gemini 2.5, is rolling out starting today, March 25, with an experimental release of Gemini 2.5 Pro, it announced in a blog post. Gemini 2.5 is Google’s “most intelligent AI model” yet, per the company. The experimental version of Gemini 2.5 Pro backs up that claim, ranking first on LMArena’s leaderboard — ahead of Grok 3, GPT-4.5, and DeepSeek R1.
Google has been ramping up its AI development, releasing the Gemini 2.0 Flash Thinking model last month. It can conduct multi-step reasoning and use certain techniques to reach a conclusion, all while the end user can trace its reasoning process in real time.
This time around, all the Gemini 2.5 models are thinking models, and are “capable of reasoning through their thoughts before responding, resulting in enhanced performance and improved accuracy.”
Gemini 2.5 Pro sits atop many of the benchmarks and leaderboards for large language models (LLMs), and Google says it achieved this feat by “combining a significantly enhanced base model with improved post-training.”
The release and development of Gemini 2.5 is intended to push Google further into the agentic era. The company says it is building thinking capabilities into all its Gemini 2.5 models moving forward so that they can solve complex problems and serve as context-aware AI agents.
Gemini 2.5 Pro also performs well in knowledge, science, and math benchmarks. It scored 18.8% on Humanity’s Last Exam without tool use, which beats all other competitors by a significant margin. Google describes this test as “a dataset designed by hundreds of subject matter experts to capture the human frontier of knowledge and reasoning.”
With native multimodality, this series of Gemini AI models can intake text, audio, images, video, and large code repositories for extra context.
Additionally, Gemini 2.5 Pro is said to perform better at coding, hence why it is shipping today in Google AI Studio. It’ll also come to Vertex AI at a later date.
Gemini 2.5 Pro pricing and availability
Gemini 2.5 Pro is available starting today for business and personal accounts with a one million token context window. It’s available in Google AI Studio and the Gemini web client and mobile apps. Currently, Gemini 2.5 Pro is exclusive to Gemini Advanced and requires a Google One AI Premium plan ($20/month).
Gemini 2.5 Pro will arrive in Vertex AI in the coming weeks, and its context window will soon expand to two million tokens.