Skip to content

GLM-Zero-Preview

A reasoning model trained with extended reinforcement learning

Visit official site

What GLM-Zero-Preview is

The model is a reasoning model trained with extended reinforcement learning to strengthen maths, code and deep reasoning.

What you can do with it

  • Work on mathematical logic and code problems
  • Follow the full reasoning process in the output

What to watch out for

  • The general ability claim is framed as no significant drop
  • A formal GLM-Zero release is described as future work

Pros & cons

✓ What we like

  • Free chat access and API calls for developers
  • Full reasoning processes are described as output

! What to watch out for

  • The general ability claim is framed as no significant drop
  • A formal GLM-Zero release is described as future work

FAQ

Which evaluations are named?

AIME 2024, MATH500 and LiveCodeBench are named.

How can developers access it?

API calls and a free trial package of 20 million tokens.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review