Reportedly: HongYun's HyperLM V2.5 Adds Native Support for Huawei Ascend 950DT and Atlas 850E SuperPoD
Reported by Pandaily · not yet confirmed by the company or a second independent outlet. We update this page when it is.
HongYun's HyperLM V2.5 now natively supports Huawei's Ascend 950DT and Atlas 850E SuperPoD, adding UnifiedBus 2.0 support, mxFP4 tuning, vNPU slicing, and prefill-decode split inference. The source announcement was published on 2026-11-04, Pandaily reported.
China context
- Outside China
- Not stated in the sources yet
- Claims
- Company-reported; not yet independently evaluated
- For builders
- Developers targeting Huawei Ascend hardware can now consider HyperLM V2.5 with native optimizations, potentially simplifying deployment on Ascend-based systems.
- For investors
- The adaptation of HyperLM V2.5 to Huawei Ascend hardware may indicate growing software support for Huawei's AI chips, which could affect the competitive landscape for AI accelerators.
HongYun's HyperLM V2.5 has been adapted to run natively on Huawei's Ascend 950DT and Atlas 850E SuperPoD. The update includes support for UnifiedBus 2.0, mxFP4 tuning, vNPU slicing, and prefill-decode split inference. This enables the model to leverage Huawei's Ascend hardware for inference workloads.
The adaptation includes UnifiedBus 2.0 support, mxFP4 tuning, vNPU slicing, and prefill-decode split inference, indicating optimization for Huawei's Ascend architecture. These features suggest improved memory bandwidth utilization, lower precision inference, and efficient distributed inference across multiple NPUs.
Huawei Ascend hardware users gain access to HyperLM V2.5 with native optimizations, potentially reducing inference costs and improving performance on Ascend-based systems. This may strengthen Huawei's position in the AI hardware market by expanding the software ecosystem for its Ascend chips.
For organizations using Huawei Ascend hardware, HyperLM V2.5 offers a model natively optimized for their infrastructure, potentially lowering integration effort and improving efficiency. This could make HyperLM V2.5 a more attractive option for deployments on Ascend-based systems.
Specific signals to watch include independent benchmarks of HyperLM V2.5 on Ascend hardware, adoption by enterprises using Huawei infrastructure, and any official announcements from HongYun or Huawei regarding certification or performance metrics.