Zhipu AI shares jump as viral Ox Alpha model revealed as GLM-5.3-Flash on Chinese chips

Zhipu AI’s GLM-5.3-Flash model, which processed 62 trillion tokens during its preview, points to China's ability to scale AI operations using domestic chips. This is a significant development for Beijing’s push to reduce reliance on US hardware. The model’s performance on OpenRouter, where it processed over 11 trillion tokens in three days and ranked first among coding systems with 10.3 trillion tokens, shows that domestic hardware can support high-demand AI applications. Our view is that the real story here is less about the model’s raw performance and more about the demonstrated capability of the 100,000 Chinese-made chips to handle such a workload. This suggests a maturing domestic supply chain for AI infrastructure. The thing to watch is how quickly this domestic chip capacity can be expanded and whether it can consistently meet the demands of future, even larger, AI models without significant performance compromises.