canirunaimodel
Z.ai · 753B (40B active) · Mixture of Experts
Z.ai flagship open-weight coding model built for long-horizon agents and 1M-token project context
Use Cases
| Quant | Bits | VRAM | Quality | Status |
|---|---|---|---|---|
| Q2_K | 2 | 241.6 GB | low | — |
| Q3_K_M | 3 | 338 GB | moderate | — |
| Q4_K_M | 4 | 386.2 GB | good | — |
| Q5_K_M | 5 | 482.6 GB | good | — |
| Q6_K | 6 | 579.1 GB | excellent | — |
| Q8_0 | 8 | 771.9 GB | excellent | — |
| F16 | 16 | 1543.3 GB | lossless | — |
About this model
Z.ai
Mixture of Experts model for chat, reasoning, code workloads.
Context
1024K
Q4 VRAM
~386.2 GB
Image saved from public model sources
GLM-5.3 on your hardware
Z.ai flagship open-weight coding model built for long-horizon agents and 1M-token project context. This page turns the Hugging Face model card into practical local-run numbers, so you can compare quantized VRAM, system RAM, and expected fit before downloading a large checkpoint.
753B total parameters with 40B active. Dense models use the whole network for each token.
Start with Q4_K_M: about 368.2 GB on disk and 386.2 GB VRAM before extra context and runtime overhead.
chat, reasoning, code
Listed as GLM-5.3 from huggingface.co/zai-org/GLM-5.3.
Want the real verdict? Pick your GPU or edit the specs on this page and compare the quant table below.
Open HF repo