Qwen2.5-Max

Qwen2.5-Max, released January 29, 2025, was Alibaba's flagship closed-weight model at the height of the post-DeepSeek-V3 Chinese model race. It is a large-scale mixture-of-experts model trained on over 20 trillion tokens with SFT and RLHF, offered as a proprietary API through Alibaba Cloud Model Studio and Qwen Chat. Alibaba claimed it outperformed DeepSeek V3, GPT-4o and Claude 3.5 Sonnet on its chosen benchmarks, including strong HumanEval coding scores; those are vendor-reported results from the launch window.

No weights were released, making it Alibaba's first sustained turn toward a frontier-closed, smaller-open strategy. It was later joined by the open Qwen2.5-VL family and superseded by the Qwen3 generation and the trillion-parameter Qwen3-Max.