GLM-5.2
753B parameter open-weight model. GLM-5.2 is a 753B MoE model requiring multi-GPU or distributed inference for full deployment. Ollama publishes a cloud-hosted entry only (`glm-5.2:cloud`, re-checked 2026-08-24) and exposes no downloadable local artifact, so this catalog records no local tag. Below data-center scale, the unsloth/GLM-5.2-GGUF 1-bit UD-IQ1 variants (~217–228 GB) via llama.cpp or Unsloth Studio with KTransformers are the only local execution path, at reduced fidelity. Not consumer-local hardware friendly.
Z.ai (Zhipu AI) · GLM
Model overview
GLM-5.2 is Z.ai's large mixture-of-experts model, aimed at coding, agentic tool-use, and very long context work rather than single-GPU home setups. At 753B total parameters it sits well outside consumer hardware territory — even the aggressive 1-bit GGUF community quantizations still need roughly 217-228 GB, so realistic local deployment means a multi-GPU workstation or distributed inference setup, not a single consumer card. It's included here for completeness and comparison, not as a practical recommendation for most local AI builders; if you're VRAM-constrained, look at the smaller open-weight models in the directory instead.
Editorial review
VRAM figures are empirical estimates. Actual usage varies by runtime, context length, and system configuration. Verify on your specific hardware before production use.
Will GLM-5.2 run on your machine?
GLM-5.2 is 753B parameters and needs 467.5 GB of VRAM at Q4_K_M — 466 GB of weights plus 1.5 GB of runtime overhead for the inference server itself.
VRAM by quantization
| Quantization | Weights | Needs (with overhead) | Quality |
|---|---|---|---|
| Q4_K_M | 466 GB | 467.5 GB | good |
| Q8_0 | 801 GB | 802.5 GB | high |
Fit on common hardware at Q4_K_M
| Hardware | Memory the model can use | System RAM | Verdict |
|---|---|---|---|
| CPU Only | None (CPU only) | 16 GB | Too large |
| RTX 4060 Laptop | 8 GB | 16 GB | Too large |
| RTX 3060 (12GB) | 12 GB | 32 GB | Too large |
| RTX 4060 Ti (16GB) | 16 GB | 32 GB | Too large |
| RTX 3090 | 24 GB | 64 GB | Too large |
| Apple Silicon (Unified Memory) 36 GB | 27 GB of 36 GB | 36 GB | Too large |
| RTX 5090 | 32 GB | 64 GB | Too large |
| Apple Silicon (Unified Memory) 48 GB | 36 GB of 48 GB | 48 GB | Too large |
| Apple Silicon (Unified Memory) 64 GB | 48 GB of 64 GB | 64 GB | Too large |
| Apple Silicon (Unified Memory) 96 GB | 72 GB of 96 GB | 96 GB | Too large |
| Apple Silicon (Unified Memory) 128 GB | 96 GB of 128 GB | 128 GB | Too large |
| Apple Silicon (Unified Memory) 192 GB | 144 GB of 192 GB | 192 GB | Too large |
Comfortable means VRAM clears the requirement by 2 GB or more. Tight means it covers the requirement with no margin. CPU offload means the model does not fit in VRAM but system RAM is at least 1.6× the weights, so it will run at reduced speed — expect roughly 1–5 tokens per second. Figures are weights plus a fixed runtime overhead and exclude KV-cache growth, which scales with context length.
Apple Silicon shares one pool of memory between the system and the GPU, so a model cannot use all of it. These rows apply the same 75% usable fraction the Compatibility Checker uses, which is why a 36 GB Mac is graded on less than 36 GB.
Need more hardware for GLM-5.2? Open the PC Builder for the MoE / frontier-style experiment tier →
This catalog has no verified local Ollama tag for this checkpoint, so fit grades do not include a local run command.
Ready to run this model locally?
Find a compatible interface in our Local AI Tools directory →