RESOLVING THE LEDGER
This is one outlet's own report from Quartz — the article as it was filed. Other outlets are covering the same event; open the full story to compare every source side by side.
See the full story · 2 sources
Z.ai confirmed Wednesday that it was behind Ox Alpha, the anonymous AI model that had been circulating on OpenRouter and topping usage charts, and released it under the name GLM-5.3-Flash. The company said every request to the model during its preview period was served on Chinese-made AI chips.
GLM-5.3-Flash carries 320 billion total parameters and 18 billion active parameters, the company said. According to Bloomberg , Z.ai set pricing at $0.15 per million tokens on the input side and $0.50 per million on the output side, putting it in the same low-cost bracket as DeepSeek. Z.ai also listed a per-task price of $0.045, which it said is roughly one-tenth the cost of comparable offerings.
The model is the first in Z.ai's GLM-5 series to support multimodal input, handling text, images, and video. On coding and agentic benchmarks, it approaches Anthropic's Claude Opus 4.8, the company said. On the Artificial Analysis Intelligence Index, the model places 10th, putting it ahead of DeepSeek V4 Pro Max, according to CNBC .
Z.ai stated that a cluster of 100,000 domestically produced chips powered every request the model received, going back to when it first appeared under the Ox Alpha name. The company did not disclose which chipmakers supplied the hardware. CNBC said it could not verify the chip claims independently. Counterpoint Senior Research Analyst Ivan Lam said the hardware is probably a mix of Huawei Ascend processors and components sourced from additional vendors. "Chinese AI model developers have continued to allocate more resources and investment toward AI servers and computing infrastructure built on domestic chips," Lam said.
Z.ai said it built a dedicated inference engine on top of the SGLang framework and achieved a 3x improvement in serving performance over its initial baseline on the same hardware, reaching efficiency and per-token cost it said are comparable to mainstream Nvidia $NVDA GPUs.
Z.ai stock climbed more than 8% in Hong Kong trading on Thursday.
Z.ai drew the Ox Alpha name from a Chinese film called "Niu Lai," which translates to "Ox Comes," Bloomberg reported. Z.ai's approach mirrored a tactic Alibaba and Xiaomi had each employed this year, floating a model anonymously so the company could gauge real-world reception before stepping forward to claim it.
Z.ai released the model weights on Wednesday on HuggingFace and made GLM-5.3-Flash available to all GLM Coding Plan subscribers, the company said. The launch comes ahead of Z.ai's first detailed earnings report covering the first half of the year, due Monday.
The Ox Alpha preview drew significant traffic on OpenRouter, where it became the most popular model of the week , with Z.ai saying all of that traffic ran on domestic Chinese chips. The company's stock has climbed more than 800% since its Hong Kong listing in January.
AIPROPX is an independent multi-source news index — we track, compare, and connect coverage from across the web into one place you won't find anywhere else.