Z.ai выпустила GLM-5.2 на чипах Huawei — модель заняла первое место в мировом рейтинге

Chinese Z.ai released the GLM-5.2 language model and immediately claimed first place in the Artificial Analysis index — one of the most authoritative independent benchmarks assessing the quality of AI systems. Moreover, the entire GLM-5 lineup was trained exclusively on Huawei Ascend 910B processors — without a single GPU from Nvidia.

Huawei Instead of Nvidia

The choice of Ascend 910B is not a forced measure due to sanctions, but a declared position of the Z.ai team. The company explicitly states that Nvidia equipment was not used in principle during training. This is an important signal: a sovereign AI stack on Chinese hardware has reached a competitive level with American products.

Z.ai is a company known in China for its series of GLM (General Language Model) models developed jointly with Tsinghua University. GLM is actively used in the corporate sector of the PRC. Now the company is for the first time claiming leadership in a global independent rating.

Ascend 910B was created by Huawei as a response to Nvidia's A100. For a long time it was perceived as a worthy but inferior alternative — especially in matrix computation speed and framework support. The GLM-5.2 result calls this perception into question. Apparently, the team bet on data quality and algorithm efficiency rather than scaling up through additional chips.

Open Code vs. Closed Flagships

While the US keeps its most powerful closed models — Fable 5 and Mythos 5 — from widespread distribution, Z.ai is moving in the opposite direction. GLM-5.2 is released with open source code: it can be downloaded and run locally, without API keys and cloud dependencies. Any company in the world can take the model, fine-tune it on their own data, and deploy it within their own perimeter.

What this gives companies and developers in practice:

• Full control over data — everything is processed within your own infrastructure
• Independence from foreign providers and possible trade restrictions
• Ability to fine-tune for specific company tasks and domains
• No token-based pricing — only the cost of your own hardware
• Deployment in an isolated perimeter without data going to third parties

Sanctions Accelerated Rather Than Stopped

Since 2022, the US has consistently tightened export controls on high-performance GPUs for China: first a ban on A100, then A800, and then H20. The strategic logic was simple: without access to the best chips, China would slow down in AI development. The GLM-5.2 result puts this logic under serious doubt.

This is not the first precedent: in early 2025, DeepSeek R1 showed that algorithmic efficiency can compensate for hardware limitations. Z.ai took the next step — proving that "sanctioned" chips can not only build competitive models but also rank first in a global rating.

Artificial Analysis is an authoritative independent benchmark aggregator, a reference point for corporate buyers and developers when choosing a platform. First place in this index converts to real commercial interest and attracts new corporate contracts.

What This Means

Sanctions strategy was designed to isolate China from AI progress — instead it triggered accelerated development of independent AI infrastructure. The industry adapted: its own chips were created, efficient algorithms developed, independent ecosystems built.

Third countries observing this race now have a real alternative — powerful, free, and without political conditions. Z.ai released a model that outperforms American analogues in benchmarks and is available to the entire world with open source code. For the global market, this is a clear signal: the era of unchallenged dominance of American AI companies is ending.