Z.ai open-sources ‘Ox Alpha’ model as GLM-5.3-Flash

Z.ai has open-sourced GLM-5.3-Flash, a large language model previously released anonymously on OpenRouter as Ox Alpha. The model processed over 20 trillion tokens in six days before its identity was revealed. Featuring a mixture-of-experts architecture with 320 billion total parameters and 18 billion active parameters, it offers a one-million-token context window and native multimodal input. GLM-5.3-Flash is ten times more cost-efficient than its predecessor and provides MIT-licensed weights for AI engineering teams.