NextFin News -- Alibaba has updated its flagship model Qwen3.8-Max, enhancing the overall performance of the model after targeted retraining in programming (Coding) and professional office collaboration (Cowork).
In the global authoritative third-party ranking CodeArena, which focuses on front-end programming capabilities (WebDev), the model improved by 22 points to a total score of 1691, surpassing other models such as Claude Opus5 and Kimi K3, and securing the top position on the overall leaderboard.
Additionally, the updated model cost-performance ranking (Pareto Frontier) on CodeArena indicates that the new version of Qwen3.8-Max has an average comprehensive cost of only $5 per million tokens, beating all other models priced above $5.
Qwen3.8-Max is Alibaba's most powerful large language model to date, with a total of 2.4 trillion parameters and support for 1 million context tokens. Currently, the new Qwen3.8-Max model has been launched on the Qianwen AI platform, providing API services, and has been promptly integrated into Qianwen Office, Qoder, and the Qianwen APP.
Explore more exclusive insights at nextfin.ai.
Insights
What is the Qwen series of large language models?
What does the 2.4 trillion parameter count signify for model capability?
How does context window size affect large model performance?
How does Qwen3.8-Max compare to competitors like Claude Opus5?
What is the current market position of Alibaba AI models globally?
How do users perceive the cost-performance ratio of Qwen3.8-Max?
What specific improvements were made in the latest Qwen3.8-Max update?
How did Qwen3.8-Max perform on the CodeArena leaderboard?
Which Alibaba products have integrated the new Qwen3.8-Max model?
How might this update impact Alibaba cloud computing business growth?
What role will coding capabilities play in future AI model competition?
How could the 1 million context token support change enterprise applications?
What challenges exist in verifying third-party benchmark rankings like CodeArena?
How does Alibaba balance model performance with inference costs?
What limitations remain for Qwen3.8-Max despite the update?
How does Qwen3.8-Max compare to Kimi K3 in programming tasks?
What distinguishes Qwen3.8-Max from previous versions of the Qwen model?
How does the 5 dollars per million tokens pricing compare to industry standards?
