GLM 4.6V needs about 72.5 GB of memory at Q4_K_M (a 70.4 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 2 quants, from 70.4 GB (Q4_K_M) to 113.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 6.2 GB.
GLM 4.6V – a 107B multimodal model with 128k token context
GLM 4.6V is a 107B parameter model that supports vision and text. It was trained with a 131072 token context window and offers native multimodal function calling. The model is released under the MIT license.
Good at
- Vision‑text reasoning
- Native multimodal function calling
- Interleaved image‑text generation
Limits
- No audio or video support
- Requires large VRAM for full model
- No official support for low‑latency local use
Best for
- Multimodal document analysis
- Vision‑driven tool integration
- Complex image‑text content creation
Who it’s for. Developers building cloud‑based multimodal agents or researchers needing high‑capacity vision‑text models. The model is suitable for high‑performance clusters rather than edge devices.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 70.4 GB | 72.5 GB | Too big | Get file → | |
| Q8_0 | 113.6 GB (3 parts) | 115.7 GB | Too big | Get file → |
For 8‑GB VRAM use the Q4_K_M quantized file (70.45 GB, ~72.5 GB memory at 8k context). For larger GPUs, the Q8_0 quantized file (113.61 GB, ~115.7 GB memory at 8k context) can be used, split into three shards.
What GPU do you need for GLM 4.6V?
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 GLM 4.6V quant should you choose?
Q4_K_M is the usual starting point for GLM 4.6V: it balances file size and quality well. Here it is a 70.4 GB file.
The Q8_0 file is 113.6 GB, about 1.6 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.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about GLM 4.6V
How much VRAM does GLM 4.6V need?
About 72.5 GB at Q4_K_M with an 8K context window. That is the 70.4 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 4.6V on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 72.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of GLM 4.6V GGUF?
It depends on the quant: Q4_K_M is 70.4 GB, Q8_0 is 113.6 GB.
What context length does GLM 4.6V support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 1.5 GB at 8K and 6.2 GB at 32K.
Where can I download GLM 4.6V GGUF?
The files are on Hugging Face in the ggml-org/GLM-4.6V-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 4.6V 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 4.6V on 8 GB VRAM?
The Q4_K_M quantized file requires about 72.5 GB memory at 8k context, so 8 GB VRAM is insufficient.
Does GLM 4.6V support audio input?
No, the model only detects vision and text abilities.
Is there a lightweight version for local deployment?
The parent model includes a 9B GLM‑4.6V‑Flash, but the GGUF files provided are for the full 107B model.