GLM-4.5-Air GGUF size and VRAM requirements
unsloth/GLM-4.5-Air-GGUF is a very large language model with 110.47 billion parameters, built on the glm4moe architecture. It is released under the mit license and has been downloaded 23,012 times.
To run unsloth/GLM-4.5-Air-GGUF locally at a 4,096-token context, its quantized versions need between 36.67 GB (Q1_0, lowest quality) and 206.86 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.
GGUF file size and memory by quantization
Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| Q1_0 | 2.77 | Low | 35.63 GB | 0.24 GB | 36.67 GB | — | Insufficient |
| IQ1_S | 2.79 | Low | 35.86 GB | 0.24 GB | 36.9 GB | — | Insufficient |
| IQ1_M | 2.9 | Low | 37.31 GB | 0.24 GB | 38.35 GB | — | Insufficient |
| IQ2_XXS | 3.09 | Low | 39.75 GB | 0.24 GB | 40.79 GB | — | Insufficient |
| IQ2_M | 3.21 | Low | 41.34 GB | 0.24 GB | 42.38 GB | — | Insufficient |
| Q2_K | 3.28 | Low | 42.21 GB | 0.24 GB | 43.25 GB | — | Insufficient |
| Q2_K_L | 3.29 | Low | 42.34 GB | 0.24 GB | 43.39 GB | — | Insufficient |
| Q2_K_XL | 3.44 | Fair | 44.19 GB | 0.24 GB | 45.23 GB | — | Insufficient |
| IQ3_XXS | 3.73 | Fair | 47.91 GB | 0.24 GB | 48.95 GB | — | Insufficient |
| Q3_K_S | 3.81 | Fair | 49.04 GB | 0.24 GB | 50.08 GB | — | Insufficient |
| Q3_K_XL | 3.97 | Fair | 51.02 GB | 0.24 GB | 52.06 GB | — | Insufficient |
| Q3_K_M | 4.14 | Fair | 53.28 GB | 0.24 GB | 54.32 GB | — | Insufficient |
| IQ4_XS | 4.38 | Good | 56.33 GB | 0.24 GB | 57.37 GB | — | Insufficient |
| IQ4_NL | 4.54 | Good | 58.4 GB | 0.24 GB | 59.44 GB | — | Insufficient |
| Q4_0 | 4.55 | Good | 58.47 GB | 0.24 GB | 59.51 GB | — | Insufficient |
| Q4_K_S | 4.85 | Good | 62.43 GB | 0.24 GB | 63.47 GB | — | Insufficient |
| Q4_K_XL | 4.9 | Good | 63.07 GB | 0.24 GB | 64.11 GB | — | Insufficient |
| Q4_1 | 5.03 | Very good | 64.64 GB | 0.24 GB | 65.68 GB | — | Insufficient |
| Q4_K_M | 5.28 | Very good | 67.96 GB | 0.24 GB | 69.0 GB | — | Insufficient |
| Q5_K_S | 5.68 | Very good | 73.03 GB | 0.24 GB | 74.07 GB | — | Insufficient |
| Q5_K_XL | 6.0 | Very good | 77.16 GB | 0.24 GB | 78.2 GB | — | Insufficient |
| Q5_K_M | 6.05 | Very good | 77.78 GB | 0.24 GB | 78.82 GB | — | Insufficient |
| Q6_K | 7.17 | Excellent | 92.21 GB | 0.24 GB | 93.25 GB | — | Insufficient |
| Q6_K_XL | 7.35 | Excellent | 94.58 GB | 0.24 GB | 95.62 GB | — | Insufficient |
| Q8_0 | 8.51 | Excellent | 109.39 GB | 0.24 GB | 110.43 GB | — | Insufficient |
| Q8_K_XL | 9.25 | Excellent | 118.97 GB | 0.24 GB | 120.01 GB | — | Insufficient |
| BF16 | 16.0 | Excellent | 205.82 GB | 0.24 GB | 206.86 GB | — | Insufficient |
KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
How much VRAM do you need to run unsloth/GLM-4.5-Air-GGUF?
You need about 45.23 GB of VRAM to run unsloth/GLM-4.5-Air-GGUF entirely on the GPU using the Q2_K_XL quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run unsloth/GLM-4.5-Air-GGUF on an 8 GB GPU?
No. unsloth/GLM-4.5-Air-GGUF does not fit on an 8 GB GPU, even with the smallest quantization and system RAM offloading.
Can I run unsloth/GLM-4.5-Air-GGUF on a 16 GB GPU?
Partially. unsloth/GLM-4.5-Air-GGUF only fits on a 16 GB GPU by offloading part of it to system RAM (with Q2_K_XL), which runs but is slower.
Can I run unsloth/GLM-4.5-Air-GGUF on a 24 GB GPU?
Partially. unsloth/GLM-4.5-Air-GGUF only fits on a 24 GB GPU by offloading part of it to system RAM (with Q4_K_M), which runs but is slower.
What is the best quantization for unsloth/GLM-4.5-Air-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.