GLM-5.3-Flash Ships Under MIT, and Z.ai Confirms It Was Ox Alpha

Z.ai released GLM-5.3-Flash on August 26, 2026. The model had spent the previous week on OpenCode and OpenRouter under the name Ox Alpha, the anonymous listing we covered on August 23; Z.ai's launch identified the two as the same model, which answers the question of who built it.

It is a mixture-of-experts model with 320B total parameters and 18B active per token, and the first natively multimodal model in the GLM-5 series. The weights are on Hugging Face at zai-org/GLM-5.3-Flash under the MIT licence.

Z.ai's pricing page lists $0.15 per 1M input tokens, $0.03 for cached input and $0.50 for output. The equivalent GLM-5.3 rates are $1.40, $0.26 and $4.40. A 50% promotion runs until 24:00 on September 9, 2026 (UTC+8), taking GLM-5.3-Flash to $0.075 / $0.015 / $0.25.

The API documentation gives a 1M-token context window and a 128K maximum output, with thinking mode, function calling, context caching and structured output. On the GLM Coding Plan the model carries three times the quota of GLM-5.3, and calls made off-peak, including all day at weekends, consume half the standard points.

Z.ai reports 84.3 on Terminal-Bench 2.1 and 63.4 on Deep-SWE, and states that the model beats GLM-5.2 across its evaluations at a tenth of the price. Those are the vendor's own numbers, measured with its own scaffold.