KA

Karnak

KarnakGGUF by mradermacherUpdated 2026-05-13
Download on Hugging Face
Parameters40.7B
Context256K
Layers64
Architectureqwen3moe
Smallest quant0.2 GB
Quick answer

Karnak needs about 26.3 GB of memory at Q4_K_M (a 24.7 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 24 quants, from 0.2 GB to 33.4 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

Karnak – a 40.7B parameter Arabic‑English LLM in GGUF format

Karnak is a 40.7B parameter model built on Qwen3‑moe architecture. It supports Arabic and English generation with an extended context window of 262144 tokens. The model is available in multiple quantizations for local deployment.

Good at

  • Arabic‑English instruction following
  • Long‑form text generation
  • Tool‑use capability

Limits

  • No vision support
  • Requires large VRAM for higher‑quality quants
  • License is Apache‑2.0

Best for

  • Content creation in Arabic and English
  • Research summarization
  • Fine‑tuning for domain‑specific tasks

Who it’s for. Developers and researchers who need a multilingual LLM that can run locally. The model is suitable for those who prefer open‑source licensing and want to fine‑tune for specific use cases.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.2 GB 1.7 GB
Fits Get file →
IQ1_S8.5 GB 10.1 GB
Too big Get file →
IQ1_M9.4 GB 11.0 GB
Too big Get file →
IQ2_XXS10.9 GB 12.5 GB
Too big Get file →
IQ2_XS12.1 GB 13.7 GB
Too big Get file →
IQ2_S12.4 GB 13.9 GB
Too big Get file →
IQ2_M13.5 GB 15.1 GB
Too big Get file →
Q2_K_S14.0 GB 15.6 GB
Too big Get file →
Q2_K15.0 GB 16.6 GB
Too big Get file →
IQ3_XXS15.8 GB 17.4 GB
Too big Get file →
IQ3_XS16.8 GB 18.3 GB
Too big Get file →
Q3_K_S17.7 GB 19.3 GB
Too big Get file →
IQ3_S17.7 GB 19.3 GB
Too big Get file →
IQ3_M18.0 GB 19.6 GB
Too big Get file →
Q3_K_M19.6 GB 21.2 GB
Too big Get file →
Q3_K_L21.2 GB 22.8 GB
Too big Get file →
IQ4_XS21.8 GB 23.4 GB
Too big Get file →
Q4_023.1 GB 24.7 GB
Too big Get file →
Q4_K_S23.2 GB 24.8 GB
Too big Get file →
Q4_K_M24.7 GB 26.3 GB
Too big Get file →
Q4_125.6 GB 27.1 GB
Too big Get file →
Q5_K_S28.1 GB 29.7 GB
Too big Get file →
Q5_K_M28.9 GB 30.5 GB
Too big Get file →
Q6_K33.4 GB 35.0 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (24.72 GB file, ~26.3 GB memory at 8 k context). For 12 GB VRAM, Q5_K_M (28.94 GB file, ~30.5 GB memory). For 16 GB VRAM, Q6_K (33.42 GB file, ~35 GB memory). These estimates are approximate.

What GPU do you need for Karnak?

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: the best quant that fits is the main file (0.2 GB file, about 1.7 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 1.7 GB in use).
  • 8 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 1.7 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ1_M (9.4 GB file, about 11.0 GB in use).
  • 16 GB VRAM: the best quant that fits is Q2_K_S (14.0 GB file, about 15.6 GB in use).
  • 24 GB VRAM: the best quant that fits is IQ4_XS (21.8 GB file, about 23.4 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (28.9 GB file, about 30.5 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (33.4 GB file, about 35.0 GB in use).

Which Karnak quant should you choose?

Q4_K_M is the usual starting point for Karnak: it balances file size and quality well. Here it is a 24.7 GB file.

If memory is tight, the smallest file is the first one in the table at 0.2 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen3moe
Parameters40.7B (approximate, mixture of experts)
Layers64
Trained context262,144 tokens
KV cache (fp16)128 KB per token

Source

Quantized bymradermacher
FamilyKarnak
Licenseapache-2.0
Downloads (HF)2,026
Last updated2026-05-13

Frequently asked questions about Karnak

How much VRAM does Karnak need?

About 26.3 GB at Q4_K_M with an 8K context window. That is the 24.7 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 Karnak on an 8 GB GPU?

Yes, with the smaller files: the 0.2 GB file needs about 1.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Karnak GGUF?

It depends on the quant: file is 0.2 GB, IQ1_S is 8.5 GB, IQ1_M is 9.4 GB, IQ2_XXS is 10.9 GB, IQ2_XS is 12.1 GB, IQ2_S is 12.4 GB and 18 more.

What context length does Karnak support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download Karnak GGUF?

The files are on Hugging Face in the mradermacher/Karnak-i1-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 Karnak 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 Karnak on 8 GB VRAM?

Yes, the Q4_K_M quant uses about 26.3 GB of VRAM at 8 k context, which fits within 8 GB when using 8 k context.

Does Karnak support vision tasks?

No, Karnak is a text‑only model with no vision capability.

Other GGUFs of Karnak-40B-v1.0

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Karnak models

All Karnak →

Other large models

All large models →