Chinese artificial intelligence company Z.ai confirmed on Wednesday that the previously anonymous Ox Alpha model, which gained viral attention last week, was an early test version of its new GLM-5.3-Flash model. The company stated in a blog post that Ox Alpha was released anonymously on OpenRouter and OpenCode on August 20 to gather user feedback before its official launch. The model operated on Chinese-made AI chips and ranked as the most-used model on OpenRouter in its first week, according to Z.ai.
The announcement followed days of speculation among developers and internet users about the model’s origins. OpenRouter described Ox Alpha as a reasoning model optimized for coding, sustained agentic work, and production workloads, while OpenCode offered it for free with near-unlimited usage capacity. The model’s weights were made publicly available on Hugging Face, allowing developers to download, modify, and build upon it.
Z.ai’s shares surged 8% in Hong Kong trading on Thursday, reflecting investor enthusiasm for the company’s latest release. The model’s low-cost design and reliance on domestic chips have been highlighted as key factors in its performance and accessibility. Rival Chinese AI firm MiniMax also saw a 3% share increase after reporting a 283% revenue surge in the first half of the year.
Technical and market context
Z.ai’s GLM-5.3-Flash is a multimodal model capable of processing text, images, and video with a context window of up to 1 million tokens. The company claims the model was tested using 100,000 Chinese-made chips to handle all online requests, though these details were not independently verified by CNBC. The model’s performance has drawn praise from international figures, including Stripe CEO Patrick Collison, who called it “very impressive” in a post on X.
The model’s name, Ox Alpha, sparked additional intrigue due to its association with the viral Chinese animated film Niu Lai (Here Comes the Ox), which became a summer meme on social media. Some users speculated the name was a playful nod to Chinese cultural references, while others analyzed technical signals that hinted at Z.ai’s involvement. Earlier this year, Z.ai tested a similar model anonymously under the name Pony Alpha, further fueling speculation.
Global and industry reactions
The release of GLM-5.3-Flash has been framed as a step forward for Chinese open-source AI models, with users on Weibo celebrating the model’s performance and affordability. One user from Beijing praised its strong performance at a low cost, while another from Chongqing noted its significance in advancing China’s position in the global developer market. A user from Sichuan highlighted Z.ai’s use of domestic chips, emphasizing the model’s efficiency on homegrown hardware.
The model’s success on OpenRouter—where it ranked 10th on the Artificial Analysis Intelligence Index, ahead of DeepSeek V4 Pro Max—has drawn comparisons to leading U.S. AI models, though these are not officially available in China due to U.S. export restrictions on advanced chips. The restrictions have accelerated China’s push for semiconductor self-sufficiency, with companies like Huawei developing alternatives to foreign chips such as those from Nvidia.
Market and competitive landscape
Z.ai’s announcement coincides with a broader trend of Chinese AI companies gaining traction in both domestic and international markets. The surge in Z.ai’s shares reflects investor confidence in the company’s ability to compete with established players, including those backed by U.S. technology. Meanwhile, MiniMax’s revenue growth underscores the rapid expansion of China’s AI sector, despite regulatory and supply chain challenges.
The model’s release also highlights the growing role of open-source AI tools in fostering innovation. By making the model’s weights publicly available, Z.ai has enabled developers worldwide to experiment with and adapt the technology, further accelerating its adoption. The company’s approach contrasts with the closed ecosystems of some Western AI firms, offering a more collaborative model for advancement.