Post Thumbnail

Google Gemini 2.5 Pro surpassed OpenAI o3 and leads on LMArena

Google updated Gemini 2.5 Pro with strong benchmark gains. Now the model surpasses the current o3 version from OpenAI. Like 2.5 Flash, this is a hybrid model that allows setting budget for thinking processes or completely turning them off. The model is already available and according to first impressions works better than early versions. It even listens when you ask it not to spam code with comments.

On LMArena arena, the new version rose by 24 Elo points compared to the previous one and now leads in all categories, surpassing o3 and Claude Opus 4. On benchmarks, the model noticeably improved and became more efficient in code, logic and exact sciences tasks.

The results are impressive. 82.2% on programming tasks, 86.4% on natural sciences questions and 21.6% on Humanity’s Last Exam test, which checks thinking and knowledge.

Developers also took into account feedback about the previous version and improved style and structure. Now the model can be more creative. They also added budgets for thinking processes for greater cost control. Unfortunately, image generation for Gemini Pro was still not added.

Autor: AIvengo
For 5 years I have been working with machine learning and artificial intelligence. And this field never ceases to amaze, inspire and interest me.

Latest News

Samsung seeks replacement for Google Gemini for Galaxy S26

Samsung Electronics, one of the leading mobile device manufacturers, is actively seeking alternatives to Google Gemini for its future Galaxy S26 lineup. The company is conducting negotiations with OpenAI and Perplexity, striving to expand the artificial intelligence ecosystem in its devices.

How language models transfer knowledge through random numbers

Have you ever wondered if numbers can store knowledge? Scientists discovered an amazing phenomenon. Language models can transfer their behavioral traits through sequences of digits that look like random noise.

Alibaba introduced Quark AI smart glasses with Snapdragon AR1 chip

Chinese tech giant Alibaba introduced its first model of Quark AI smart glasses at the World Conference on Artificial Intelligence in Shanghai.

Why advanced AI models confuse themselves during long reasoning

You give a complex task to a smart person and expect that the longer they think, the more accurate the answer will be. Logical, right? That's exactly how we're used to thinking about artificial intelligence work too. But new research from Anthropic shows that reality is much more interesting.

Z.AI introduced GLM-4.5 with 355 billion parameters and open source

Meet the new technological heavyweight! Z.AI company introduced the open language model GLM-4.5, which is ready to challenge Western giants not only with capabilities but also with accessibility.