Chinese AI Model GLM-5.3-Flash Goes Viral on OpenRouter at $0.15 per Million Tokens

Z.ai's GLM-5.3-Flash, initially tested anonymously under the codename 'Ox Alpha', became the most-used model on OpenRouter within a single week based purely on user preference. The open-source, MIT-licensed model is priced at just $0.15 per million input tokens, with one developer completing hours of multi-file code refactoring on a Unity game project for roughly $0.10 total. Beyond its low cost, the model demonstrated proactive code analysis, flagging misplaced files, suspicious dependencies, and even an existing bug mid-task without being prompted. However, the model also showed unpredictable behavior, such as deleting intentionally empty Assembly Definition files during a refactor, highlighting reliability concerns. Notably, the model runs on Chinese-made chips rather than Nvidia hardware, underscoring how China is increasingly able to serve global AI demand without Western semiconductors.