Melusine 103b needs about 49.3 GB of memory at Q3_K_S (a 44.8 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 5 quants, from 38.2 GB (Q2_K) to 46.4 GB (IQ3_M).
Melusine 103b is a 103B-parameter Melusine model in GGUF format, quantized by mradermacher. Its trained context length is 32K tokens. The repository has 5 quant files, from 38.2 GB (Q2_K) to 46.4 GB (IQ3_M).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 38.2 GB | 42.8 GB | Too big | Get file → | |
| IQ3_XS | 42.5 GB | 47.0 GB | Too big | Get file → | |
| Q3_K_S | 44.8 GB | 49.3 GB | Too big | Get file → | |
| IQ3_S | 44.9 GB | 49.4 GB | Too big | Get file → | |
| IQ3_M | 46.4 GB | 51.0 GB | Too big | Get file → |
What GPU do you need for Melusine 103b?
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: the best quant that fits is IQ3_XS (42.5 GB file, about 47.0 GB in use).
Which Melusine 103b quant should you choose?
Q3_K_S (44.8 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
If memory is tight, the smallest file is Q2_K at 38.2 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 Melusine 103b
How much VRAM does Melusine 103b need?
About 49.3 GB at Q3_K_S with an 8K context window. That is the 44.8 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 Melusine 103b on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 42.8 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Melusine 103b GGUF?
It depends on the quant: Q2_K is 38.2 GB, IQ3_XS is 42.5 GB, Q3_K_S is 44.8 GB, IQ3_S is 44.9 GB, IQ3_M is 46.4 GB.
What context length does Melusine 103b support?
The GGUF header lists a trained context length of 32,764 tokens (32K). The KV cache grows with the context you use: about 4.0 GB at 8K and 16.1 GB at 32K.
Where can I download Melusine 103b GGUF?
The files are on Hugging Face in the mradermacher/Melusine_103b-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.