Recently, the Tongyi Lingma platform has upgraded its model selection feature by adding the full-powered DeepSeek V3 and R1 models. Additionally, the Qwen2.5-Max model has been officially launched. The Qwen2.5-Max is pre-trained on over 20 trillion tokens and utilizes a specially designed post-training approach for in-depth optimization.

Users can now select their desired model versions from the drop-down menu within the input boxes of Tongyi Lingma's intelligent Q&A and AI Programmer interfaces. This update offers users a broader range of model options to suit various application scenarios and requirements.

In multiple benchmark tests, Qwen2.5-Max has demonstrated exceptional performance. It outperformed other industry-leading models such as DeepSeek V3, GPT-4o (although direct access to its base model was not available), and Claude-3.5-Sonnet in evaluations like Arena-Hard, LiveBench, LiveCodeBench, and GPQA-Diamond. Moreover, Qwen2.5-Max achieved competitive results in assessments like MMLU-Pro.

When compared with open-source models, Qwen2.5-Max continues to shine. Due to the inability to directly compare closed-source base models, Qwen2.5-Max was benchmarked against notable models like DeepSeek V3 (a leading open-source MoE model), Llama-3.1-405B (the largest open-source dense model), and Qwen2.5-72B (another top-tier open-source dense model). The results show that Qwen2.5-Max holds a significant advantage in these comparisons.

Notably, in the latest blind test rankings for large models published by the third-party benchmarking platform Chatbot Arena, Qwen2.5-Max scored 1,332 points, securing the seventh position globally. It also stands out as one of the top non-reasoning Chinese large models. In specific skill tests like mathematics and programming, Qwen2.5-Max ranked first, while in hard prompt tests, it ranked second.

The official evaluation from ChatBot Arena highlighted that Qwen2.5-Max exhibits strong capabilities across multiple domains, especially in technical fields like programming and mathematics. This further solidifies Qwen2.5-Max’s leading position in the high-performance large model landscape.

Currently, the Tongyi Lingma platform has integrated the Qwen2.5-Max model, allowing users to experience its powerful coding abilities by downloading the Tongyi Lingma plugin. This update will undoubtedly provide users with a more efficient and intelligent experience.