cerebras/GLM-4.7-REAP-218B-A32B needs about 135.6 GB of memory at Q4_K_M (a 132.0 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 53.0 GB (TQ1_0) to 436.9 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 12.3 GB.
GLM-4.7-REAP-218B-A32B: 218B sparse MoE model compressed 40% from 355B
GLM-4.7-REAP-218B-A32B is a memory‑efficient compressed variant of GLM‑4.7. It uses REAP pruning to reduce parameters to 218B while keeping 32B active per token. The model supports a context length of 202,752 tokens and is licensed under MIT.
Good at
- Large context reasoning
- Sparse MoE efficiency
- Near‑lossless performance on code and function tasks
Limits
- Requires high VRAM for full precision
- No vision support
- Limited to 8,000 token context for memory estimates
Best for
- Local deployment on high‑VRAM GPUs
- Academic research needing large context
- Fine‑tuning sparse MoE models
Who it’s for. Developers and researchers who need a large‑context language model but have limited GPU memory. The model is suitable for local inference and fine‑tuning on commodity hardware with sufficient VRAM.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 53.0 GB | 56.6 GB | Too big | Get file → | |
| IQ1_S | 60.6 GB (2 parts) | 64.2 GB | Too big | Get file → | |
| IQ1_M | 67.3 GB (2 parts) | 70.8 GB | Too big | Get file → | |
| IQ2_XXS | 72.1 GB (2 parts) | 75.7 GB | Too big | Get file → | |
| IQ2_M | 76.0 GB (2 parts) | 79.6 GB | Too big | Get file → | |
| Q2_K | 80.4 GB (2 parts) | 84.0 GB | Too big | Get file → | |
| Q2_K_L | 80.6 GB (2 parts) | 84.2 GB | Too big | Get file → | |
| Q2_K_XL | 83.8 GB (2 parts) | 87.4 GB | Too big | Get file → | |
| IQ3_XXS | 89.5 GB (2 parts) | 93.1 GB | Too big | Get file → | |
| Q3_K_S | 94.8 GB (2 parts) | 98.4 GB | Too big | Get file → | |
| Q3_K_XL | 97.8 GB (2 parts) | 101.4 GB | Too big | Get file → | |
| Q3_K_M | 104.8 GB (3 parts) | 108.4 GB | Too big | Get file → | |
| IQ4_XS | 116.9 GB (3 parts) | 120.5 GB | Too big | Get file → | |
| IQ4_NL | 123.6 GB (3 parts) | 127.2 GB | Too big | Get file → | |
| Q4_0 | 124.0 GB (3 parts) | 127.5 GB | Too big | Get file → | |
| Q4_K_S | 124.3 GB (3 parts) | 127.9 GB | Too big | Get file → | |
| Q4_K_XL | 124.9 GB (3 parts) | 128.5 GB | Too big | Get file → | |
| Q4_K_M | 132.0 GB (3 parts) | 135.6 GB | Too big | Get file → | |
| Q4_1 | 137.0 GB (3 parts) | 140.6 GB | Too big | Get file → | |
| Q5_K_S | 150.7 GB (4 parts) | 154.3 GB | Too big | Get file → | |
| Q5_K_XL | 154.8 GB (4 parts) | 158.4 GB | Too big | Get file → | |
| Q5_K_M | 155.1 GB (4 parts) | 158.6 GB | Too big | Get file → | |
| Q6_K | 179.5 GB (4 parts) | 183.1 GB | Too big | Get file → | |
| Q6_K_XL | 184.5 GB (4 parts) | 188.1 GB | Too big | Get file → | |
| Q8_0 | 232.2 GB (5 parts) | 235.8 GB | Too big | Get file → | |
| Q8_K_XL | 241.8 GB (5 parts) | 245.4 GB | Too big | Get file → | |
| BF16 | 436.9 GB (9 parts) | 440.5 GB | Too big | Get file → |
For 8‑GB VRAM, use Q4_K_S (file size 124.34 GB, memory 127.9 GB at 8k context). For 16‑GB VRAM, Q5_K_S (150.68 GB, 154.3 GB). For 24‑GB VRAM, Q6_K (179.47 GB, 183.1 GB). Estimates are approximate.
What GPU do you need for cerebras/GLM-4.7-REAP-218B-A32B?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which cerebras/GLM-4.7-REAP-218B-A32B quant should you choose?
Q4_K_M is the usual starting point for cerebras/GLM-4.7-REAP-218B-A32B: it balances file size and quality well. Here it is a 132.0 GB file.
The Q8_0 file is 232.2 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is TQ1_0 at 53.0 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about cerebras/GLM-4.7-REAP-218B-A32B
How much VRAM does cerebras/GLM-4.7-REAP-218B-A32B need?
About 135.6 GB at Q4_K_M with an 8K context window. That is the 132.0 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run cerebras/GLM-4.7-REAP-218B-A32B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 56.6 GB, so you would need a larger GPU or CPU offloading.
What is the file size of cerebras/GLM-4.7-REAP-218B-A32B GGUF?
It depends on the quant: TQ1_0 is 53.0 GB, IQ1_S is 60.6 GB, IQ1_M is 67.3 GB, IQ2_XXS is 72.1 GB, IQ2_M is 76.0 GB, Q2_K is 80.4 GB and 21 more.
What context length does cerebras/GLM-4.7-REAP-218B-A32B support?
The GGUF header lists a trained context length of 202,752 tokens (198K). The KV cache grows with the context you use: about 3.1 GB at 8K and 12.3 GB at 32K.
Where can I download cerebras/GLM-4.7-REAP-218B-A32B GGUF?
The files are on Hugging Face in the unsloth/GLM-4.7-REAP-218B-A32B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does cerebras/GLM-4.7-REAP-218B-A32B use?
Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run GLM-4.7-REAP-218B-A32B on 8 GB VRAM?
The smallest quant for 8 GB VRAM is Q4_K_S, which requires about 127.9 GB of memory at 8k context. It is not feasible on 8 GB VRAM without further quantization.
Does the model support vision tasks?
No, the model does not include vision capabilities.
What license does the model use?
The model is released under the MIT license.