DA

Dark-Miqu-103B

DarkGGUF by backyardaiUpdated 2024-05-22
Download on Hugging Face
Parameters103B
Context32K
Layers120
Architecturellama
Smallest quant21.7 GB
Quick answer

Dark-Miqu-103B needs about 66.4 GB of memory at Q4_K_M (a 61.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 23 quants, from 21.7 GB (IQ1_S) to 206.4 GB (F16).

Dark-Miqu-103B: 103B‑parameter dark fantasy model with 32k context

Dark-Miqu-103B is a 103‑billion‑parameter LLM designed for dark and grimdark fantasy writing. It supports a 32764‑token context window and is available in multiple GGUF quantizations for local deployment.

Good at

  • Dark fantasy prose
  • Long‑form creative writing
  • Low‑positivity tone

Limits

  • No vision support
  • License type: other
  • Requires 8k context for memory estimates

Best for

  • Writing grimdark stories
  • Generating dark world lore
  • Creative writing prompts

Who it’s for. Developers and writers who need a large, locally‑hosted model for dark fantasy content. It is suitable for use with llama.cpp‑compatible applications such as Backyard AI.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S21.7 GB 26.2 GB
Too big Get file →
IQ1_M23.8 GB 28.3 GB
Too big Get file →
IQ2_XXS27.3 GB 31.8 GB
Too big Get file →
IQ2_XS30.4 GB 34.9 GB
Too big Get file →
IQ2_S31.9 GB 36.4 GB
Too big Get file →
IQ2_M34.7 GB 39.2 GB
Too big Get file →
Q2_K_S35.1 GB 39.6 GB
Too big Get file →
Q2_K37.9 GB 42.4 GB
Too big Get file →
IQ3_XXS39.6 GB 44.1 GB
Too big Get file →
IQ3_XS42.2 GB 46.7 GB
Too big Get file →
Q3_K_S44.5 GB 49.0 GB
Too big Get file →
IQ3_S44.6 GB 49.1 GB
Too big Get file →
IQ3_M46.1 GB 50.7 GB
Too big Get file →
Q3_K_M49.6 GB 54.1 GB
Too big Get file →
Q3_K_L54.1 GB (2 parts) 58.6 GB
Too big Get file →
IQ4_XS55.1 GB (2 parts) 59.6 GB
Too big Get file →
Q4_K_S58.6 GB (2 parts) 63.1 GB
Too big Get file →
Q4_K_M61.9 GB (2 parts) 66.4 GB
Too big Get file →
Q5_K_S71.0 GB (2 parts) 75.5 GB
Too big Get file →
Q5_K_M72.9 GB (2 parts) 77.5 GB
Too big Get file →
Q6_K84.7 GB (2 parts) 89.2 GB
Too big Get file →
Q8_0109.7 GB (3 parts) 114.2 GB
Too big Get file →
F16206.4 GB (5 parts) 210.9 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (21.67 GB file, 26.2 GB memory). For 16 GB VRAM, IQ2_M (34.7 GB file, 39.2 GB memory) or Q2_K (37.92 GB file, 42.4 GB memory). For 32 GB VRAM, Q4_K_M (61.89 GB file, 66.4 GB memory) or Q5_K_M (72.93 GB file, 77.5 GB memory).

What GPU do you need for Dark-Miqu-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: the best quant that fits is IQ2_XXS (27.3 GB file, about 31.8 GB in use).
  • 48 GB VRAM: the best quant that fits is IQ3_XS (42.2 GB file, about 46.7 GB in use).

Which Dark-Miqu-103B quant should you choose?

Q4_K_M is the usual starting point for Dark-Miqu-103B: it balances file size and quality well. Here it is a 61.9 GB file.

The Q8_0 file is 109.7 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 IQ1_S at 21.7 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

Architecturellama
Parameters103B
Layers120
Trained context32,764 tokens
KV cache (fp16)480 KB per token

Source

Quantized bybackyardai
FamilyDark
Licenseother
Downloads (HF)1,004
Last updated2024-05-22

Frequently asked questions about Dark-Miqu-103B

How much VRAM does Dark-Miqu-103B need?

About 66.4 GB at Q4_K_M with an 8K context window. That is the 61.9 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 Dark-Miqu-103B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 26.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Dark-Miqu-103B GGUF?

It depends on the quant: IQ1_S is 21.7 GB, IQ1_M is 23.8 GB, IQ2_XXS is 27.3 GB, IQ2_XS is 30.4 GB, IQ2_S is 31.9 GB, IQ2_M is 34.7 GB and 17 more.

What context length does Dark-Miqu-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 Dark-Miqu-103B GGUF?

The files are on Hugging Face in the backyardai/Dark-Miqu-103B-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 Dark-Miqu-103B use?

Hugging Face lists the license as other. 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 Dark-Miqu-103B on 8 GB VRAM?

Yes, using the IQ1_S quantization which requires 26.2 GB of memory at 8k context, but you can reduce context size to fit your VRAM.

Does Dark-Miqu-103B support vision?

No, the model has no vision capabilities.

What is the context length of Dark-Miqu-103B?

The trained context length is 32764 tokens.

Similar models

Browse all →

More from backyardai

All by backyardai →

More Dark models

All Dark →

Other large models

All large models →