Reportedly: Ascend, MetaX, Biren and Moore Threads Announce Day-0 Support for StepFun's Step 5 Preview
Reported by Pandaily · not yet confirmed by the company or a second independent outlet. We update this page when it is.
Four domestic chip platforms announced day-0 support for StepFun's Step 5 Preview on its open-weights release day. Huawei said its first 1,024-card Atlas 950 SuperPoD went live alongside it, Pandaily reported.
China context
- Original name
- 阶跃星辰
- Outside China
- Open weights · pandaily.com
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers outside China can access Step 5 Preview weights and potentially run it on supported Chinese chips if available, but must verify export and licensing constraints.
- For investors
- The coordinated day-0 support from four Chinese chip vendors signals a maturing domestic AI stack, which may reduce NVIDIA's market share in China for inference workloads.
On the day StepFun released open weights for Step 5 Preview, four Chinese chip platforms—Huawei Ascend, MetaX, Biren Technology, and Moore Threads—announced support for the model. Huawei also stated that its first 1,024-card Atlas 950 SuperPoD went live at the same time. Step 5 Preview is a 600B total parameter model with 27B active parameters and a 1M context window, featuring a 92-layer design, Sparse GQA, and long-horizon agent training.
Step 5 Preview uses a sparse mixture-of-experts architecture with 600B total parameters and 27B active parameters, supporting a 1M context window. The model employs Sparse GQA and a 92-layer design, with training focused on long-horizon agent tasks. Day-0 support from four domestic chip vendors suggests the model is optimized for Chinese AI accelerators, potentially reducing reliance on NVIDIA hardware.
Chinese AI chip vendors gain immediate compatibility with a major open-weights model, strengthening their position in the domestic inference market and reducing dependence on foreign GPUs for StepFun's model users.
Enterprises in China can deploy Step 5 Preview on domestic AI accelerators from day one, potentially lowering infrastructure costs and meeting data-sovereignty requirements.
Watch for independent benchmarks of Step 5 Preview on domestic chips versus NVIDIA GPUs, and whether other Chinese model providers follow with day-0 multi-vendor chip support.