GLM API pricing for every Z.ai model, newest first

GLM 5.3 Prime is the high speed build of GLM 5.3, with 1.5 to 2 times the throughput. Z.ai doesn't sell it on its own API, so the row shows Alibaba, at $2.80 input and $8.80 output per million. GLM 5.3 FlashX Legal costs no more per output token than GLM 5.3 FlashX, at $1.25 per million, so the legal training is free on output. Test it on your own documents all the same.

A cached token on GLM 5.3 FlashX Legal costs $0.075 per million, 80% off the normal input rate. No batch price is listed for GLM 5.3 right now.

Zhipu publishes open weights for many GLM models, so you can also run them yourself or buy them from other hosts. GLM 5.3 FlashX Legal reads up to 1.05M tokens, the largest window in the family.

Z.ai also sells the GLM Coding Plan, a monthly subscription for coding tools with Lite, Pro and Max tiers. It gives you a pool of prompts that resets every five hours, not a token bill, so it isn't priced here.

For general work, start with GLM 5.3 and run GLM 5.3 Flash, the lowest cost model Z.ai sells for the example request, on the same prompts. Use it wherever it passes your checks. Keep GLM 5.3 FlashX Legal for contracts and other legal text, and reach for FlashX or Prime only where users wait on the answer.

If most of your use is a coding tool, compare the Coding Plan with your token bill before you pick.