Expired60% confidenceFactExact time
Zhipu's model is reportedly trained on Huawei Ascend 910D chips, with the 910D targeting a benchmark against NVIDIA's H100.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/6/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 10/4/2026(expired)
Sources
GLM-4 Open Source Strikes Back: The AI War Expands to Memory and Space Computing
bilibili智灵Shaw7/3/2026
Related Claims
UnverifiedHuawei's Ascend chips are based on its self-developed DaVinci architecture, with the 910D targeting performance on par with NVIDIA's H100.82% similarVerifiedHuawei Ascend's hardware product line includes the Ascend 910 series training chips and the Ascend 310 series inference chips.79% similarUnverifiedThe Ascend 950 is Huawei's latest-generation AI training chip, and the 128-card super-node solution refers to connecting 128 Ascend 950 chips via high-speed interconnects to form a single logical training unit.78% similarUnverifiedHuawei's Ascend series is currently the domestic alternative with overall performance closest to NVIDIA's H100 and has been deployed at scale in training clusters at multiple domestic large model companies.72% similarUnverifiedAfter restrictions were imposed, Chinese labs primarily relied on downgraded A800 chips and domestically produced Huawei Ascend chips for model training.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/114154API
curl https://kongchang.com/api/v1/knowledge/claims/114154MCP
get_claim(id=114154)