Google Unveils Gemini 2.5, Its Next AI Reasoning Model

(CLO) On March 26, Google officially announced Gemini 2.5, a new generation of AI models capable of pausing to "think" before giving an answer.

This is an important step in the race to develop AI with logical thinking, helping to improve accuracy and performance in complex tasks such as programming and mathematics.

Google officially announced Gemini 2.5. Photo: Google

Kicking off the new generation, Google is launching Gemini 2.5 Pro Experimental, a multimodal AI model that the company claims is its most powerful yet. This version will be available on the Google AI Studio developer platform and the Gemini app for subscribers of the $20/month Gemini Advanced plan.

Google also confirmed that all of their new AI models in the future will integrate this thinking ability.

Since OpenAI launched o1 – the first thinking AI model in September 2024, leading technology companies such as Anthropic, DeepSeek, Google and xAI have entered the race to develop AI with stronger thinking capabilities. These models use more computational resources, analyze more carefully before giving answers, to improve accuracy and reduce errors.

According to experts, thinking AI will be the core element of autonomous AI agents - tools that can perform many tasks without human intervention. However, the downside of these models is that the operating costs are significantly higher than traditional AI models.

Google previously experimented with thinking AI with a version of Gemini in December 2024. But Gemini 2.5 marks Google's most serious attempt to surpass OpenAI's "o" line of models.

Google claims that the Gemini 2.5 Pro outperforms its previous AI models and many other competitors on the market. In particular, it is designed to excel at web programming and agent coding.

Google’s launch of Gemini 2.5 signals a shift in AI from simple data processing to deeper reasoning. Models like Gemini 2.5 not only improve performance in programming and mathematics, but could pave the way for more advanced autonomous AI systems in the future.

However, the race for thinking AI is far from over. OpenAI, Anthropic, xAI, and other competitors are certainly not standing still, and the battle between the AI ​​giants will only get more intense in 2025.

Cao Phong (according to TheVerge, MSB, TC)

Source: https://www.congluan.vn/google-ra-mat-gemini-25-mo-hinh-ly-luan-ai-tiep-theo-post340083.html