Austral 32B GLM4 Winton needs about 20.7 GB of memory at Q4_K_M (a 19.7 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 26 quants, from 9.9 GB (IQ2_XS) to 65.1 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 2.0 GB.
32.6B parameter GLM4 model with 32k context
Austral 32B GLM4 Winton is a large language model with 32.6B parameters and a trained context of 32768 tokens. It is released under the Apache‑2.0 license and supports a wide range of quantizations for llama.cpp usage.
Good at
- High‑capacity text generation
- Long‑form context handling
- Versatile quantization options
Limits
- No vision support
- Requires 8k context for memory estimates
- Large model size may need high VRAM
Best for
- Chatbots needing long context
- Research prototypes
- Content generation pipelines
Who it’s for. Developers and researchers who need a large GLM4 model with extensive context and flexible quantization. The model is suitable for local deployment using llama.cpp or compatible frameworks.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XS | 9.9 GB | 10.9 GB | Too big | Get file → | |
| IQ2_S | 10.4 GB | 11.4 GB | Too big | Get file → | |
| IQ2_M | 11.3 GB | 12.3 GB | Too big | Get file → | |
| Q2_K | 12.3 GB | 13.3 GB | Too big | Get file → | |
| IQ3_XXS | 12.8 GB | 13.8 GB | Too big | Get file → | |
| Q2_K_L | 13.2 GB | 14.2 GB | Too big | Get file → | |
| IQ3_XS | 13.7 GB | 14.7 GB | Too big | Get file → | |
| Q3_K_S | 14.4 GB | 15.4 GB | Too big | Get file → | |
| IQ3_M | 14.8 GB | 15.8 GB | Too big | Get file → | |
| Q3_K_M | 15.9 GB | 16.9 GB | Too big | Get file → | |
| Q3_K_L | 17.2 GB | 18.2 GB | Too big | Get file → | |
| IQ4_XS | 17.6 GB | 18.6 GB | Too big | Get file → | |
| Q3_K_XL | 18.0 GB | 19.0 GB | Too big | Get file → | |
| IQ4_NL | 18.6 GB | 19.6 GB | Too big | Get file → | |
| Q4_0 | 18.6 GB | 19.6 GB | Too big | Get file → | |
| Q4_K_S | 18.7 GB | 19.7 GB | Too big | Get file → | |
| Q4_K_M | 19.7 GB | 20.7 GB | Too big | Get file → | |
| Q4_K_L | 20.4 GB | 21.4 GB | Too big | Get file → | |
| Q4_1 | 20.5 GB | 21.6 GB | Too big | Get file → | |
| Q5_K_S | 22.5 GB | 23.5 GB | Too big | Get file → | |
| Q5_K_M | 23.1 GB | 24.1 GB | Too big | Get file → | |
| Q5_K_L | 23.7 GB | 24.7 GB | Too big | Get file → | |
| Q6_K | 26.7 GB | 27.7 GB | Too big | Get file → | |
| Q6_K_L | 27.2 GB | 28.2 GB | Too big | Get file → | |
| Q8_0 | 34.6 GB | 35.6 GB | Too big | Get file → | |
| BF16 | 65.1 GB (2 parts) | 66.2 GB | Too big | Get file → |
For 8 GB VRAM, use IQ2_XS (9.9 GB file, 10.9 GB memory at 8k context). For 16 GB VRAM, Q4_K_M (19.68 GB file, 20.7 GB memory) or Q5_K_S (22.53 GB file, 23.5 GB memory) are suitable. For 32 GB VRAM, Q6_K (26.73 GB file, 27.7 GB memory) or Q8_0 (34.61 GB file, 35.6 GB memory) provide higher quality.
What GPU do you need for Austral 32B GLM4 Winton?
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: the best quant that fits is IQ2_S (10.4 GB file, about 11.4 GB in use).
- 16 GB VRAM: the best quant that fits is IQ3_M (14.8 GB file, about 15.8 GB in use).
- 24 GB VRAM: the best quant that fits is Q5_K_S (22.5 GB file, about 23.5 GB in use).
- 32 GB VRAM: the best quant that fits is Q6_K_L (27.2 GB file, about 28.2 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (34.6 GB file, about 35.6 GB in use).
Which Austral 32B GLM4 Winton quant should you choose?
Q4_K_M is the usual starting point for Austral 32B GLM4 Winton: it balances file size and quality well. Here it is a 19.7 GB file.
The Q8_0 file is 34.6 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 IQ2_XS at 9.9 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Austral 32B GLM4 Winton
How much VRAM does Austral 32B GLM4 Winton need?
About 20.7 GB at Q4_K_M with an 8K context window. That is the 19.7 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 Austral 32B GLM4 Winton on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 10.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Austral 32B GLM4 Winton GGUF?
It depends on the quant: IQ2_XS is 9.9 GB, IQ2_S is 10.4 GB, IQ2_M is 11.3 GB, Q2_K is 12.3 GB, IQ3_XXS is 12.8 GB, Q2_K_L is 13.2 GB and 20 more.
What context length does Austral 32B GLM4 Winton support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.5 GB at 8K and 2.0 GB at 32K.
Where can I download Austral 32B GLM4 Winton GGUF?
The files are on Hugging Face in the bartowski/Delta-Vector_Austral-32B-GLM4-Winton-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 Austral 32B GLM4 Winton use?
Hugging Face lists the license as apache-2.0. 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?
Yes, using the IQ2_XS quantization which requires about 10.9 GB of memory at 8k context.
Does this model support vision tasks?
No, the model does not include vision capabilities.
What license applies to this model?
The model is released under the Apache‑2.0 license.