Z.ai Runs GLM-5.3-Flash On 100,000 Chinese Chips. It Tops OpenRouter. The Hardware Independence Is The Point.
Z.ai deployed GLM-5.3-Flash, served exclusively on 100,000 China-made chips, under the stealth codename Ox Alpha on August 20. The model ranked first by usage on OpenRouter globally within a week. Hong Kong-listed shares rose over 8 percent on Thursday. Rival MiniMax reported nearly 300 percent first-half revenue growth year over year, with shares up about 3 percent.
This demonstrates the principle of compute sovereignty. When you control the silicon, you control the deployment timeline and the cost ceiling. The mechanism here is hardware-software co-optimization: a model trained and served on domestic chips removes a single point of geopolitical failure. Everyone watching the export restriction saga should understand that constraints breed parallel infrastructure.
Z.ai, also known as Zhipu, deployed GLM-5.3-Flash on 100,000 Chinese chips and topped OpenRouter usage globally. MiniMax posted nearly 300 percent first-half revenue growth year over year.
- Go to openrouter.ai and create a free account to access its model leaderboard.
- Search for GLM-5.3-Flash in the model directory and note its usage rank and pricing.
- Send a basic prompt through OpenRouter's playground to compare its output quality against a model like Claude or GPT. Expected outcome: you observe firsthand that a model running entirely on non-Nvidia hardware can compete at the top of a global usage board.