Pricing

Pay for the compute and storage you use.

GLM 5.3 usage rates in US dollars
Usage	Rate	Unit
Input	$12	per million tokens
Cached input	$2.50	per million verified cached tokens
Output	$30	per million tokens, including reasoning
Training	$36	per million training tokens
Memory hosting	$0.30	per GB-month, prorated

$10 minimum. No subscription or automatic top-up.Try now

Learning has no per-call fee.

Each call uses input, output, and training tokens to prepare, learn, and check your knowledge. You set a spending limit before it starts.

A saved memory includes both serving and training weights. Hosting is billed by their size and retention time, with an initial 24-hour window. Version management and routing are included.

See how usage is calculated

Supported modelGLM 5.3More coming.