GLM-5.3-Flash lands under MIT at $0.15 per million input tokens

Z.ai released GLM-5.3-Flash, a 320-billion-parameter model that activates only 18 billion parameters per token, under the permissive MIT license, with weights, API access and a coding subscription plan live the same day according to its own thread. It handles text and images natively and accepts up to 1 million tokens of context in one request. Z.ai's listed pricing is $0.15 per million input tokens and $0.50 per million output tokens, with cached input at $0.03, cheap enough to make self-hosting or high-volume API use practical for a large context window without frontier pricing.

Z.ai's announcement says the model performs on par with Claude Opus 4.8 on its own coding benchmark, which is a vendor's measurement of its own model, not an outcome checked by an outside party. The same thread reveals GLM-5.3-Flash previously ran under the name Ox Alpha, and SemiAnalysis reported separately that Ox Alpha's roughly 100 trillion tokens of daily free serving ran entirely on domestic Chinese chips rather than Nvidia hardware, at efficiency it says was comparable to mainstream Nvidia accelerators, a relevant data point for anyone assessing how dependent large-scale Chinese model serving still is on Nvidia.