Run unsloth/GLM-5.2-GGUF locally
unsloth/GLM-5.2-GGUF is a very large language model with 753.86 billion parameters, built on the glm-dsa architecture. It is released under the mit license and has been downloaded 392,857 times.
To run unsloth/GLM-5.2-GGUF locally at a 4,096-token context, its quantized versions need between 202.75 GB (IQ1_S, lowest quality) and 1405.34 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.
All quantizations
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| IQ1_S | 2.3 | Very low | 201.83 GB | 0.12 GB | 202.75 GB | — | Insufficient |
| IQ1_M | 2.42 | Very low | 212.8 GB | 0.12 GB | 213.72 GB | — | Insufficient |
| IQ2_XXS | 2.53 | Very low | 222.08 GB | 0.12 GB | 223.0 GB | — | Insufficient |
| IQ2_M | 2.53 | Very low | 222.19 GB | 0.12 GB | 223.11 GB | — | Insufficient |
| Q2_K_XL | 2.69 | Low | 236.44 GB | 0.12 GB | 237.36 GB | — | Insufficient |
| IQ3_XXS | 2.99 | Low | 262.34 GB | 0.12 GB | 263.26 GB | — | Insufficient |
| IQ3_S | 3.28 | Low | 287.44 GB | 0.12 GB | 288.36 GB | — | Insufficient |
| Q3_K_M | 3.64 | Fair | 319.2 GB | 0.12 GB | 320.11 GB | — | Insufficient |
| Q3_K_XL | 3.64 | Fair | 319.41 GB | 0.12 GB | 320.33 GB | — | Insufficient |
| IQ4_XS | 3.88 | Fair | 340.22 GB | 0.12 GB | 341.14 GB | — | Insufficient |
| IQ4_NL | 3.95 | Fair | 347.07 GB | 0.12 GB | 347.98 GB | — | Insufficient |
| Q4_K_S | 4.63 | Good | 406.46 GB | 0.12 GB | 407.37 GB | — | Insufficient |
| Q4_K_M | 4.94 | Good | 433.83 GB | 0.12 GB | 434.75 GB | — | Insufficient |
| Q4_K_XL | 4.96 | Good | 435.2 GB | 0.12 GB | 436.11 GB | — | Insufficient |
| Q5_K_S | 5.6 | Very good | 491.05 GB | 0.12 GB | 491.96 GB | — | Insufficient |
| Q5_K_M | 5.95 | Very good | 522.31 GB | 0.12 GB | 523.23 GB | — | Insufficient |
| Q5_K_XL | 5.97 | Very good | 523.84 GB | 0.12 GB | 524.75 GB | — | Insufficient |
| Q6_K | 6.64 | Excellent | 582.88 GB | 0.12 GB | 583.79 GB | — | Insufficient |
| Q6_K_XL | 7.26 | Excellent | 637.37 GB | 0.12 GB | 638.28 GB | — | Insufficient |
| Q8_0 | 8.5 | Excellent | 746.32 GB | 0.12 GB | 747.24 GB | — | Insufficient |
| Q8_K_XL | 8.7 | Excellent | 763.41 GB | 0.12 GB | 764.32 GB | — | Insufficient |
| BF16 | 16.0 | Excellent | 1404.42 GB | 0.12 GB | 1405.34 GB | — | Insufficient |
KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
Can I run unsloth/GLM-5.2-GGUF on an 8 GB GPU?
No. unsloth/GLM-5.2-GGUF does not fit on an 8 GB GPU, even with the smallest quantization and system RAM offloading.
Can I run unsloth/GLM-5.2-GGUF on a 16 GB GPU?
No. unsloth/GLM-5.2-GGUF does not fit on a 16 GB GPU, even with the smallest quantization and system RAM offloading.
Can I run unsloth/GLM-5.2-GGUF on a 24 GB GPU?
No. unsloth/GLM-5.2-GGUF does not fit on a 24 GB GPU, even with the smallest quantization and system RAM offloading.
What is the best quantization for unsloth/GLM-5.2-GGUF?
If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.