GLM 5.3 Flash dealignai orcarouter merge needs about 220.8 GB of memory at IQ4_XS (a 170.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 2 quants, from 170.9 GB (IQ4_XS) to 341.0 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 197.6 GB.
GLM 5.3 Flash – orthogonal merge of two uncensored models
GLM 5.3 Flash is a 320 billion‑parameter model built by merging two independent uncensored edits of the same base. It supports vision and has a maximum context of 1 048 576 tokens. The model is released under the MIT license.
Good at
- Vision tasks
- Large context handling
- Uncensored responses
Limits
- No fine‑tuning required
- Requires 8 k context for memory estimates
- No explicit language list
Best for
- Multimodal content generation
- Long‑form text with images
- Research on uncensored LLMs
Who it’s for. Researchers and developers needing a large, vision‑capable, uncensored language model. The model is suitable for experimentation and deployment where the MIT license permits.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ4_XS | 170.9 GB | 220.8 GB | Too big | Get file → | |
| Q8_0 | 341.0 GB | 390.9 GB | Too big | Get file → |
For 8 k context, the IQ4_XS quant uses about 220.8 GB of memory and the Q8_0 quant uses about 390.9 GB. Choose IQ4_XS for tighter memory budgets and Q8_0 for higher precision.
What GPU do you need for GLM 5.3 Flash dealignai orcarouter merge?
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 5.3 Flash dealignai orcarouter merge quant should you choose?
IQ4_XS (170.9 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about GLM 5.3 Flash dealignai orcarouter merge
How much VRAM does GLM 5.3 Flash dealignai orcarouter merge need?
About 220.8 GB at IQ4_XS with an 8K context window. That is the 170.9 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 5.3 Flash dealignai orcarouter merge on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 220.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of GLM 5.3 Flash dealignai orcarouter merge GGUF?
It depends on the quant: IQ4_XS is 170.9 GB, Q8_0 is 341.0 GB.
What context length does GLM 5.3 Flash dealignai orcarouter merge support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 49.4 GB at 8K and 197.6 GB at 32K.
Where can I download GLM 5.3 Flash dealignai orcarouter merge GGUF?
The files are on Hugging Face in the Cypherfox/GLM-5.3-Flash-dealignai-orcarouter-merge 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 5.3 Flash dealignai orcarouter merge 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?
No. The smallest quant requires 220.8 GB of memory at 8 k context, far exceeding 8 GB VRAM.
Does the model support text only?
No, it includes vision capabilities and requires appropriate image handling parameters.
Is the model fully uncensored?
Yes, it has had safety‑refusal behavior substantially removed, but users must comply with the MIT license.