Qwen2.5-Max: Exploring the Capabilities of a Large-Scale MoE Model

Alibaba Qwen released Qwen2.5-Max, a large language model based on the Mixture-of-Experts architecture, trained on over 20 trillion tokens. The model outperforms DeepSeek V3 on several benchmarks and is available through Alibaba Cloud APIs and Qwen Chat.

Alibaba/Qwen

Alibaba Qwen Unveils QwQ-Max-Preview: A New Model with Deep Reasoning and Multi-Domain Potential

Alibaba's Qwen team has released QwQ-Max-Preview, a preview model built on Qwen2.5-Max, excelling in deep reasoning, mathematics, coding, and Agent tasks. Future plans include an official open-source release under Apache 2.0, a dedicated Qwen Chat app, and smaller reasoning models like QwQ-32B for local deployment.

Alibaba/Qwen