Google unveiled its most powerful generative artificial intelligence model yet, one that opens a new front in the competitive AI race as it performs mostly head-and-shoulders above its staunchest rivals in industry benchmarks. Gemini 2.5, the latest from Google’s family of generative AI models with the same name, is its “most intelligent” model and outperforms top-of-the-line models from OpenAI, Anthropic, Grok and DeepSeek “by a significant margin,” Google DeepMind Chief Technology Officer Koray Kavukcuoglu wrote in a Wednesday (March 26) blog post. That means enterprises deploying generative AI models can tap a more powerful tool in the Google Cloud arsenal to build their custom apps.
Like Google’s other Gemini models, Gemini 2.5 is natively multimodal, meaning it can analyze and understand text, audio, video, images and code — capabilities built in from the ground up, not bolted on. Gemini 2.5 also offers a context window of 1 million tokens (about 750,000 English words), so it can accept very long prompts, a feature matched only by Alibaba on some of its Qwen generative AI models. “The context window is incredibly important for the AI race,” Ilia Badeev, head of data science at Trevolution Group, told PYMNTS. “The length of the context is one of the most crucial parameters for the practical use of [AI models].” To read more about what Ilia Badeev said, click here.
We provide communication in written form only
pr@dyninno.com
Phone
+44 7391 796792
v.veltmane@dyninno.com
j.kondratovica@dynatech.lv
Phone
+37 120 65 5702
+37 129 61 3971
karnika.bahuguna@dyninno.in
i.cerevco@datapro.md
Phone
+37 379 420400
j.carreno@dyninno.com
Phone
+571 314 49 00053
We do not comment on anything which might negatively impact our business, our partnerships, our employees, or our competitors
We are happy to provide information at any time of day or night but ask you to understand that we require up to two hours to prepare any statement
