Chinese AI startup DeepSeek on Friday released the preview version of its latest model, DeepSeek-V4, and open-sourced it simultaneously, 15 months after the launch of its previous-generation DeepSeek-V3 model.

The new release continues DeepSeek's reputation for strong cost-performance while adding support for domestically developed chips, including Huawei's Ascend processors, in a move seen as strengthening China's independent AI computing ecosystem and challenging Nvidia's dominance in the sector.

Huawei later announced that, through close coordination between model and chip technologies, its full Ascend supernode lineup now supports the DeepSeek V4 series. Ascend A2, A3 and 950 products are compatible with both DeepSeek V4-Flash and DeepSeek V4-Pro models.

According to DeepSeek, the V4 series expands context length from 128K tokens to 1 million tokens, representing nearly a tenfold increase in processing capacity and enabling more advanced long-context tasks.

The company said DeepSeek V4-Pro has reached the highest level among currently available open-source models and performs on par with leading closed-source systems in several areas. It highlighted major gains in agent capabilities, allowing the model to better handle multi-step and tool-based tasks.