Z.ai Confirms Ox Alpha Is GLM-5.3-Flash and Releases Weights Under MIT License

My take: The mystery is solved: Ox Alpha is GLM-5.3-Flash, the model Z.ai launched in stealth mode on OpenRouter last week. It now has a name, weights available on Hugging Face under an MIT license, and official API pricing.

I need to be honest about the benchmarks Z.ai published: 63.4% on DeepSWE was measured by the same company that built the model. That doesn't mean the numbers are wrong, but the rule still applies: a benchmark run by the manufacturer is best read with some judgment until independent evaluations confirm it — especially since last week an external researcher found performance closer to the middle of the current model pack.

What is concrete and important: the architecture is Mixture of Experts with 320 billion total parameters and 18 billion active per token, it accepts text, images, and video, supports up to one million token context, and the MIT license makes it free for commercial use with no restrictions. Any technical team can deploy it, fine-tune it, and build on top of it.

If you have a technical team, this is the kind of model worth evaluating in your own environment, with your own type of task. Having a fast process for evaluating new tools is a real advantage right now.