GLM 4.6 REAP 268B A32B Derestricted needs about 165.8 GB of memory at Q4_K_M (a 162.3 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 22 quants, from 0.5 GB to 220.7 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 12.3 GB.
GLM 4.6 REAP 268B A32B Derestricted – a large, mixture‑of‑experts language model
GLM 4.6 REAP 268B A32B Derestricted is a 268‑billion‑parameter model with 32 billion active experts. It was modified with Norm‑Preserving Biprojected Abliteration to remove refusal behavior while preserving reasoning performance. The model is released under the MIT license.
Good at
- Large‑scale reasoning
- Multi‑expert inference
- Uncensored text generation
Limits
- No vision support
- Requires large VRAM for full precision
- Limited language coverage
Best for
- Research on uncensored LLM behavior
- Enterprise reasoning tasks
- Custom fine‑tuning of large models
Who it’s for. Researchers and developers who need a high‑performance, uncensored language model and have access to sufficient GPU memory. The model is suitable for advanced NLP projects that can accommodate its size and resource demands.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.5 GB | 4.1 GB | Fits | Get file → | |
| IQ1_S | 54.3 GB | 57.9 GB | Too big | Get file → | |
| IQ1_M | 60.3 GB | 63.9 GB | Too big | Get file → | |
| IQ2_XXS | 70.3 GB | 73.9 GB | Too big | Get file → | |
| IQ2_XS | 78.5 GB | 82.1 GB | Too big | Get file → | |
| IQ2_S | 79.8 GB | 83.4 GB | Too big | Get file → | |
| IQ2_M | 87.9 GB | 91.5 GB | Too big | Get file → | |
| Q2_K_S | 90.7 GB | 94.3 GB | Too big | Get file → | |
| Q2_K | 98.1 GB | 101.7 GB | Too big | Get file → | |
| IQ3_XXS | 103.4 GB | 106.9 GB | Too big | Get file → | |
| Q3_K_S | 116.0 GB | 119.6 GB | Too big | Get file → | |
| IQ3_S | 116.1 GB | 119.7 GB | Too big | Get file → | |
| IQ3_M | 117.9 GB | 121.5 GB | Too big | Get file → | |
| Q3_K_M | 128.4 GB | 132.0 GB | Too big | Get file → | |
| Q3_K_L | 139.5 GB | 143.1 GB | Too big | Get file → | |
| IQ4_XS | 143.3 GB | 146.9 GB | Too big | Get file → | |
| Q4_0 | 152.1 GB | 155.7 GB | Too big | Get file → | |
| Q4_K_S | 152.6 GB | 156.2 GB | Too big | Get file → | |
| Q4_K_M | 162.3 GB | 165.8 GB | Too big | Get file → | |
| Q5_K_S | 185.1 GB | 188.7 GB | Too big | Get file → | |
| Q5_K_M | 190.6 GB | 194.2 GB | Too big | Get file → | |
| Q6_K | 220.7 GB | 224.3 GB | Too big | Get file → |
For 8 GB VRAM, the IQ1_S quant (54.28 GB file, 57.9 GB memory at 8 k context) is the smallest viable option. If 16 GB VRAM is available, IQ2_XS (78.55 GB file, 82.1 GB memory) offers a balance of size and quality. For 24 GB VRAM or more, Q4_K_M (162.26 GB file, 165.8 GB memory) provides higher fidelity while staying within typical GPU limits.
What GPU do you need for GLM 4.6 REAP 268B A32B Derestricted?
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: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 4.1 GB in use).
Which GLM 4.6 REAP 268B A32B Derestricted quant should you choose?
Q4_K_M is the usual starting point for GLM 4.6 REAP 268B A32B Derestricted: it balances file size and quality well. Here it is a 162.3 GB file.
If memory is tight, the smallest file is the first one in the table at 0.5 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about GLM 4.6 REAP 268B A32B Derestricted
How much VRAM does GLM 4.6 REAP 268B A32B Derestricted need?
About 165.8 GB at Q4_K_M with an 8K context window. That is the 162.3 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 GLM 4.6 REAP 268B A32B Derestricted on an 8 GB GPU?
Yes, with the smaller files: the 0.5 GB file needs about 4.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of GLM 4.6 REAP 268B A32B Derestricted GGUF?
It depends on the quant: file is 0.5 GB, IQ1_S is 54.3 GB, IQ1_M is 60.3 GB, IQ2_XXS is 70.3 GB, IQ2_XS is 78.5 GB, IQ2_S is 79.8 GB and 16 more.
What context length does GLM 4.6 REAP 268B A32B Derestricted 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 GLM 4.6 REAP 268B A32B Derestricted GGUF?
The files are on Hugging Face in the mradermacher/GLM-4.6-REAP-268B-A32B-Derestricted-i1-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 GLM 4.6 REAP 268B A32B Derestricted 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 this model on 8 GB VRAM?
Only the smallest quant, IQ1_S, may fit, but it still requires around 57.9 GB of memory at 8 k context, so 8 GB VRAM is insufficient.
Does the model support image inputs?
No, the model has no vision capabilities.
What license applies to this model?
The model is released under the MIT license.