Chinese AI laboratory Zhipu AI has released GLM-5.1 under the MIT license — one of the most permissive open-source licenses available — making it free for any use including commercial deployment without restrictions.
Scale and architecture
GLM-5.1 is a 744-billion parameter Mixture-of-Experts model with 40 billion active parameters per forward pass. This architecture delivers performance comparable to much larger dense models while keeping inference costs manageable. The model supports a 200,000 token context window.
Benchmark performance
On standard evaluations including MMLU, HumanEval and MATH, GLM-5.1 matches or exceeds GPT-4-class performance across most categories. Coding and mathematical reasoning are reported as particular strengths, with the model scoring above 85% on competitive programming benchmarks.
Why this matters
The fully open MIT release is significant because it places no restrictions on how the model is used, modified or redistributed. Previous large open models — including Meta's Llama family — carry custom licenses with various restrictions. GLM-5.1's MIT license makes it the most freely available frontier-scale model to date.
The weights are available for download on Hugging Face. Zhipu AI also operates a commercial API for teams that prefer managed inference.