Alibaba (Qwen)
External APIs
Alibaba serves the Qwen range through Model Studio. Qwen3.8 Max Prime costs $4 and $12 per million, and the weights of several Qwen models are published.
Key points : Qwen3.8 Max Prime · Open weights available · Data policy not located.
Models served : the range opens with Qwen3.8 Max Prime, released on 23 September 2026, at $4 and $12 per million for a 1,000,000-token context; Qwen3.8 Max 0902 at $2 and $6, and Qwen3.8 Omni Flash at $0.15 and $0.47. Several models in the range have published weights, including Qwen3.8 27B.
Where the data goes
As of 18 October 2026 we have not located a data security page specific to Model Studio in the English documentation. The published disclosure page describes the training corpus of the Qwen models, not the fate of customer data. The absence may come from our search address rather than from the editor, and this page will be revised if an official page is identified.
Retention and training
The English documentation we went through gives no retention period for inputs. It says nothing either about their use for training. Since the weights of several models in the range are published, it is better to serve them yourself than to assume a favourable policy.
When to use it from a sovereign platform
For a non-sensitive workload where value for money matters, the Max range compares well. For sensitive data, the existence of published Qwen weights makes on-premises execution preferable to an API call whose data policy is not documented to us.