Stealth AI Model Ox Alpha Emerges with 1 Million Token Context Window

A new AI model called Ox Alpha has emerged on OpenRouter's API platform, boasting an impressive 1,048,576-token context window. This allows it to process vast amounts of information in a single prompt, making it particularly well-suited for long-horizon software engineering tasks.

The model's anonymous debut has sent the developer community into detective mode, with many pointing towards China's Zhipu AI, now known as Z.ai, as the likely creator. Tokenizer fingerprinting and API signature matching have provided strong evidence linking Ox Alpha to the GLM-5.3 family.

Early testers have noted the model's seamless integration into coding agents and editors, with initial benchmarks suggesting a respectable throughput of 29 tokens per second. The provider has stated that no prompts will be used for training during the one-week preview period, which carries a price tag of exactly zero dollars.