Zhipu's new model, 490 tokens/s!

What happened: Zhipu released a GLM-5.3-Highspeed model in Zcode, reported by user @leo114119 and available to subscribers at random in Zcode 3.15.1. Key numbers: 490 tokens/s reported speed; GLM-5.3-FlashX peaked at 200 tokens/s; GLM-5.3-Flash about 60 tokens/s on Z.ai API; 1M token context window; 131,072 max output tokens; low, high, max thinking levels; GLM-5.1-highspeed at 400 tokens/s in May. Why it matters: The reported speed exceeds Zhipu's prior highspeed and FlashX models, and users speculate the model may relate to the anonymous OpenCode model Omen Alpha. Coverage: 12 outlets in 24h

Zhipu released a GLM-5.3-Highspeed model in Zcode, reported by user @leo114119 and available to subscribers at random in Zcode 3.15.1.

The reported speed exceeds Zhipu's prior highspeed and FlashX models, and users speculate the model may relate to the anonymous OpenCode model Omen Alpha.