GLM 5.1 is a large-scale Mixture-of-Experts (MoE) language model and a successor with improved capabilities and performance, designed for long-context reasoning, coding, and agentic workflows.
This repository holds Vultr's GGUF quantizations of it. They are hybrid quantizations: rather than applying one precision to the whole network, each layer's precision is chosen from a measured importance score, so the bits go where they change the output most.
| Attribute | Value |
|---|---|
| Base model | zai-org/GLM-5.1 |
| Architecture | glm-dsa, Mixture-of-Experts |
| Parameters | 744B total |
| Active parameters | 40B |
| Layers | 78 |
| Hidden size | 6,144 |
| Attention heads | 64 |
| Experts | 256, top-8 selection |
| Context length | ~202K |
| Vocabulary | 154,880 |
| Predecessor | GLM 5 |
| License | MIT |
| Attribute | Value |
|---|---|
| Base model | zai-org/GLM-5.1 |
| Architecture | glm-dsa, Mixture-of-Experts |
| Parameters | 744B total |
| Active parameters | 40B |
| Layers | 78 |
| Hidden size | 6,144 |
| Attention heads | 64 |
| Experts | 256, top-8 selection |
| Context length | ~202K |
| Vocabulary | 154,880 |
| Predecessor | GLM 5 |
| License | MIT |
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| GLM-5.1-Q5_K_M-*.gguf (4 parts) | Q5_K_M | 579.51 GB | 0.0190 |
| GLM-5.1-Q4_K_M-*.gguf (4 parts) | Q4_K_M | 492.95 GB | 0.0320 |
| GLM-5.1-IQ3_M-*.gguf (3 parts) | IQ3_M | 356.54 GB | 0.1503 |
| GLM-5.1-IQ3_S-*.gguf (3 parts) | IQ3_S | 352.53 GB | 0.1622 |
| File | Type | Size | KLD vs bf16 |
|---|---|---|---|
| GLM-5.1-Q5_K_M-*.gguf (4 parts) | Q5_K_M | 579.51 GB | 0.0190 |
| GLM-5.1-Q4_K_M-*.gguf (4 parts) | Q4_K_M | 492.95 GB | 0.0320 |
| GLM-5.1-IQ3_M-*.gguf (3 parts) | IQ3_M | 356.54 GB | 0.1503 |
| GLM-5.1-IQ3_S-*.gguf (3 parts) | IQ3_S | 352.53 GB | 0.1622 |
KL divergence measures how far a quantized model's next-token probabilities drift from those of the original bf16 model, averaged over every token of the evaluation text. The lower the value, the closer the quantized model is to the original; zero would mean identical output.
Every variant in this repository was evaluated against the original bf16 model for perplexity, KL divergence, top-token agreement, HellaSwag and WinoGrande, under identical conditions. The full results are maintained on the model's page in the Vultr Model Library:
https://docs.vultr.com/models/zai-glm-51/glm-5.1-fp8?tab=quantized
That page is the reference for these files. The summary above is an excerpt from it.
llama-server -hf vultr/GLM-5.1-GGUF:Q4_K_M
The model's chat template is embedded in every file, so no extra prompt formatting is needed.
The repository also includes mtp-GLM-5.1-Q8_0.gguf, the model's multi-token prediction (MTP) head, for optional speculative decoding with any quant above:
llama-server -hf vultr/GLM-5.1-GGUF:Q4_K_M --spec-type draft-mtp
These files are derived from zai-org/GLM-5.1 and are distributed under the same license, MIT. The model is the work of Z.ai; Vultr produced the quantizations.
GLM-5.1-IQ3_S-00001-of-00003.gguf 149 GBGLM-5.1-IQ3_S-00002-of-00003.gguf 149 GBGLM-5.1-IQ3_S-00003-of-00003.gguf 54 GBGLM-5.1-IQ3_M-00001-of-00003.gguf 150 GBGLM-5.1-IQ3_M-00002-of-00003.gguf 149 GBGLM-5.1-IQ3_M-00003-of-00003.gguf 58 GBGLM-5.1-Q4_K_M-00001-of-00004.gguf 149 GBGLM-5.1-Q4_K_M-00002-of-00004.gguf 148 GBGLM-5.1-Q4_K_M-00003-of-00004.gguf 148 GBGLM-5.1-Q4_K_M-00004-of-00004.gguf 48 GBGLM-5.1-Q5_K_M-00002-of-00004.gguf 150 GBGLM-5.1-Q5_K_M-00003-of-00004.gguf 150 GBGLM-5.1-Q5_K_M-00001-of-00004.gguf 150 GBGLM-5.1-Q5_K_M-00004-of-00004.gguf 130 GBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo vultr/GLM-5.1-GGUF ./model-folderzai-org/GLM-5.1