Zhipu's GLM-5.3-Highspeed Appears in ZCode Gray Release at ~490 Tokens/s

Zhipu AI has begun a limited gray release of GLM-5.3-Highspeed inside its ZCode coding tool version 3.15.1, available only to randomly selected subscribers. A user measured generation at roughly 490 tokens per second, versus about 200 tokens/s for GLM-5.3-FlashX and around 60 tokens/s for GLM-5.3-Flash. Zhipu has not officially announced the model or its throughput.

• GLM-5.3-Highspeed surfaced in ZCode 3.15.1 as a gray release for randomly chosen subscribers
• A user measured about 490 tokens/s, meaning 10,000 tokens in roughly 20 seconds
• For comparison: GLM-5.3-FlashX peaks near 200 tokens/s and GLM-5.3-Flash around 60 tokens/s
• Pi's model catalog lists a 1M-token context window and up to 131,072 output tokens