GLM-5.2 was Zhipu AI's (Z.ai) open-weight, text-only flagship, a 753B-parameter model with adjustable thinking effort. it was retired in favour of GLM-5.3, a post-training update of the same base model at the same price.
pick it if
you needed open MIT-licensed weights for long-horizon agent and coding work
you reused long prompts: cache hits billed well below the input rate
skip it if
you are choosing today; GLM-5.3 replaced it at the same price
GLM-5.3 is Z.ai's open-weight, text-only reasoning model for coding and long-horizon tasks. it keeps GLM-5.2's base model and price, with its gains coming from post-training, and reads up to 1M tokens of context.
pick it if
you want open weights for coding and long-horizon tasks
you work with long inputs: 1M context, cache hits at $0.26 per 1M
you need JSON structured output from a reasoning model
skip it if
you need image input; GLM-5.3 Flash reads images for far less
you want answers without reasoning; thinking cannot be switched off
faq
which is cheaper, GLM-5.2 or GLM-5.3?
GLM-5.2 and GLM-5.3 list the same input and output prices. on a coding-agent session (2M tokens in, 60,000 out) GLM-5.2 costs $3.06 and GLM-5.3 $3.06.
try either model
switch between models inside the same octomind session.