Google releases ‘most intelligent model to date,’ Gemini 2.5 Pro

Just a few months after the launch of Gemini 2.0 and the Priesteek ascending, Google has announced its “most intelligent model” so far, Gemini 2.5, is able to think with better performance and accuracy.

Gemini 2.5 comes three months after the most intelligent Google version, Gemini 2.0 that has provided use and job use. This new model is available as Gueini 2.5 Pro (experimental) on the Google AI studio and for GEMINI applicants on the Gemini chat interface. It will be available on VERTEX AI soon.

“AI is smarter and more able to think.”

“Now, with Gemini 2.5, we have achieved a new level of performance by combining a greatly improved basic model with training improvement,” Kavukcuoglu books. “To move forward, we build these capabilities to think directly in all of our models, so that they can deal with more complicated problems and support more capable agents.”

More context and understanding

Like Gemini 2.0 and Gemini 2.0 Flash Thinking, Gemini 2.5 Pro “thinks” before responding. The new model can deal with multimedia input from text, sound, photos, videos and large data collections. Gemini 2.5 Pro can also understand the entire code for coding projects.

Gemini 2.5 Pro provides some of the largest context windows available for experimental models on Gemini. It charges with a symbolic context window of one million million, but it will expand to two million symbols soon. Google Ai Studio Logan Kilpatrick Product Manager has been published on X that Gemini 2.5 Pro is “the first experimental model with higher rate + bills”.

Google plans to launch prices for Gemini 2.5 models soon.

The performance of coding and improving thinking

Google said the model leads to advanced measurement tests. The company said that Gemini 2.5 Pro “leads to the standards of the match and science such as GPQA and AIME 2025” said Kavukcuoglu that the model also recorded “the latest 18.8 % through models without using the tool in the last humanitarian exam,” a data set aimed at capturing human knowledge and thinking.

Gemini 2.5 Pro also strongly works on coding tasks and a better record than Gemini 2.0 in specific criteria. Google noticed the new model “outperforming visually convincing web applications and programming instructions applications, as well as converting and editing software.”

]

A more competitive market

Gemini 2.5 Pro is enters to Fathering Model Fray in a largely changed environment from Gemini 2.0 in December. Deepsek-R1 from Deepseek showed that strong models can perform well in a small part of the cost and cost calculation. Moreover, Deepseek showed that open source models can compete with more LLMS for sources, such as Openai’s O1 and O3.

Besides Deepseek models, Google must compete with thinking models in Openai. Although the latest Openaii model was GPT-4.5-not the thinking model-the company is not expected to develop more thinking models soon.

Gemini 2.5 is The new Google II model this month. In March, the company released the latest version of the small language model, GMMA 3Which presented a symbolic context model 128,000 and was best to use in the devices on the movement.