Alibaba Group Holding LtdAlibaba launches a cheaper, more efficient AI model, enhancing its competitive position and potentially boosting cloud adoption.

Alibaba Group Holding Ltd. has introduced a new AI model, Qwen3.8-Flash-Next, a 125 billion-parameter multimodal model that promises higher performance at a significantly lower cost, offering an early look at the architecture of Qwen4. Training this model uses only about one-ninth of the resources required for Qwen3.7-Plus, yet it delivers better performance in coding and office tasks. The model is competitive with newer models from Anthropic and DeepSeek, and Alibaba is releasing the model weights for developers to download and modify. Qwen will be available via QwenCloud starting at $0.16 per million input tokens and $0.47 per million output tokens. As Alibaba invests heavily in full-stack AI infrastructure, investors are watching whether lower-cost models will accelerate adoption and boost cloud growth.
Alibaba Group Holding LtdAlibaba launches a cheaper, more efficient AI model, enhancing its competitive position and potentially boosting cloud adoption.