Zixuan Li announced the launch of a high-speed variant of the GLM Flash model for developers and enterprise users. The glm-5.3-flashx model provides throughput of up to 200 tokens/s, carrying a price 2.5x higher than the base model on both the API and Coding Plan. General API users have immediate access to the new version, though those seeking access via the Coding Plan must submit an application.

The model is now available in the Hermes Agent through the Nous Portal and OpenRouter. This iteration focuses on delivery speed for large language model responses, targeting users who require faster token output than standard Flash versions provide for throughput-heavy AI tasks.