Yesterday, Alibaba introduced a new series of AI models Qwen2.5-1M, and today it announced an even more powerful version – Qwen2.5-Max. This model surpasses both DeepSeek and GPT-4o in its capabilities.
What's special about Qwen2.5-Max
- A huge amount of training data – the model was pre-trained on over 20 trillion tokens and uses state-of-the-art post-training methods.
- The MoE (Mixture-of-Experts) architecture provides high efficiency at scaling.
- Development of thinking and logic — the model successfully copes with complex questions of the university test level, can write high-quality code and even develop games.
- Analysis of multimedia content — In addition to text, Qwen2.5-Max is capable of processing images and videos.
Additional benefits
Full compatibility with the OpenAI API, which makes integration with other services even easier.
You can try Qwen2.5-Max right now — just select this model from the list on the official website.