Has Google created the smartest AI? New Gemini 2.5 outperformed all competitors in tests — and is available for free

Google introduced Gemini 2.5 is a new family of AI models capable of reasoning.

Since September 2024, when OpenAI launched the first AI reasoning model, the technology industry was in a hurry to equal or surpass its capabilities with its own developments — similar Anthropic, DeepSeek, Google and xAI, and differ from standard ones in that they use additional computing power and time to verify facts, and also run a reasoning process before giving an answer.

Gemini 2.5 Pro scored 68.6% in the Aider Polyglot test and 63.8% in the SWE-bench Verified (both of which test programming and code editing capabilities), outperforming alternatives from OpenAI and DeepSeek, but slightly behind Anthropic in the second. On the «Last Examination of Humanity», a multimodal test consisting of thousands of questions related to math, humanities, and sciences, Gemini 2.5 Pro scored 18.8% — just as well as most of the competitors’ flagship models.

In the demo video, you can see how 2.5 Pro uses reasoning to program a video game based on a single hint:

Gemini 2.5 Pro comes with a context window of 1 million tokens — meaning it can accept approximately 750,000 words at a time (more than the entire book series «Lord of the Rings»), but this is expected to double soon.

The Gemini 2.5 Pro Experimental multimodal model is already available for free at Google AI Studio and in the Gemini app for Advanced subscribers for $20/month.