JU

Juwel Beryl

JuwelGGUF by mradermacherUpdated 2026-07-30
Download on Hugging Face
Parameters40.6B
Context256K
Layers80
Architectureqwen3vl
Smallest quant0.0 GB
Quick answer

Juwel Beryl needs about 27.6 GB of memory at Q4_K_M (a 24.4 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 24 quants, from 0.0 GB to 33.3 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 10.7 GB.

Juwel Beryl – 40.6B parameter vision‑language model

Juwel Beryl is a 40.6B parameter qwen3vl architecture with 262144 trained context tokens. It supports vision and tool use. The model is released under apache‑2.0 and has no safety layer.

Good at

  • Vision‑language tasks
  • Tool integration
  • Large context handling

Limits

  • No safety layer – requires user filtering
  • No explicit language list
  • Requires 8k context for memory estimates

Best for

  • Image‑to‑text generation
  • Vision‑based question answering
  • Tool‑enabled workflows

Who it’s for. Developers building vision‑language applications who need a large context model and are prepared to implement their own safety measures. The model is suitable for users with access to sufficient GPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 3.2 GB
Fits Get file →
IQ1_S9.0 GB 12.1 GB
Too big Get file →
IQ1_M9.8 GB 12.9 GB
Too big Get file →
IQ2_XXS11.1 GB 14.3 GB
Too big Get file →
IQ2_XS12.2 GB 15.4 GB
Too big Get file →
IQ2_S12.9 GB 16.1 GB
Too big Get file →
IQ2_M14.0 GB 17.2 GB
Too big Get file →
Q2_K_S14.1 GB 17.3 GB
Too big Get file →
Q2_K15.2 GB 18.4 GB
Too big Get file →
IQ3_XXS15.8 GB 19.0 GB
Too big Get file →
IQ3_XS16.9 GB 20.1 GB
Too big Get file →
Q3_K_S17.7 GB 20.9 GB
Too big Get file →
IQ3_S17.8 GB 21.0 GB
Too big Get file →
IQ3_M18.4 GB 21.6 GB
Too big Get file →
Q3_K_M19.7 GB 22.9 GB
Too big Get file →
Q3_K_L21.4 GB 24.6 GB
Too big Get file →
IQ4_XS21.8 GB 25.0 GB
Too big Get file →
Q4_023.1 GB 26.3 GB
Too big Get file →
Q4_K_S23.2 GB 26.4 GB
Too big Get file →
Q4_K_M24.4 GB 27.6 GB
Too big Get file →
Q4_125.5 GB 28.7 GB
Too big Get file →
Q5_K_S28.0 GB 31.2 GB
Too big Get file →
Q5_K_M28.7 GB 31.9 GB
Too big Get file →
Q6_K33.3 GB 36.5 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (8.96 GB file, 12.1 GB memory at 8k context). For 16 GB VRAM, IQ2_XS (12.24 GB file, 15.4 GB memory). For 24 GB VRAM, Q4_K_M (24.43 GB file, 27.6 GB memory).

What GPU do you need for Juwel Beryl?

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 (0.0 GB file, about 3.2 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 3.2 GB in use).
  • 8 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 3.2 GB in use).
  • 12 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 3.2 GB in use).
  • 16 GB VRAM: the best quant that fits is IQ2_XS (12.2 GB file, about 15.4 GB in use).
  • 24 GB VRAM: the best quant that fits is Q3_K_M (19.7 GB file, about 22.9 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (28.7 GB file, about 31.9 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (33.3 GB file, about 36.5 GB in use).

Which Juwel Beryl quant should you choose?

Q4_K_M is the usual starting point for Juwel Beryl: it balances file size and quality well. Here it is a 24.4 GB file.

If memory is tight, the smallest file is the first one in the table at 0.0 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

Architectureqwen3vl
Parameters40.6B
Layers80
Trained context262,144 tokens
KV cache (fp16)320 KB per token

Source

Quantized bymradermacher
FamilyJuwel
Licenseapache-2.0
Downloads (HF)595
Last updated2026-07-30

Frequently asked questions about Juwel Beryl

How much VRAM does Juwel Beryl need?

About 27.6 GB at Q4_K_M with an 8K context window. That is the 24.4 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 Juwel Beryl on an 8 GB GPU?

Yes, with the smaller files: the 0.0 GB file needs about 3.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Juwel Beryl GGUF?

It depends on the quant: file is 0.0 GB, IQ1_S is 9.0 GB, IQ1_M is 9.8 GB, IQ2_XXS is 11.1 GB, IQ2_XS is 12.2 GB, IQ2_S is 12.9 GB and 18 more.

What context length does Juwel Beryl support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.7 GB at 8K and 10.7 GB at 32K.

Where can I download Juwel Beryl GGUF?

The files are on Hugging Face in the mradermacher/juwel-beryl-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 Juwel Beryl 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 Juwel Beryl on 8 GB VRAM?

Use the IQ1_S quant, which is 8.96 GB in size and requires about 12.1 GB of memory at 8k context. It may fit with careful memory management.

Does Juwel Beryl support text generation only?

No, it supports vision‑language tasks and tool use. It is not a pure text generation model.

Is there a safety layer included?

No, safety refusals have been removed. Users must add their own filtering and oversight.

Other GGUFs of juwel-beryl

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Juwel models

All Juwel →

Other large models

All large models →