OR

Rys Saber

OrnsteinGGUF by GestaltLabsUpdated 2026-04-18
Download on Hugging Face
Parameters35.5B
Context256K
Layers41
Architectureqwen35moe
Smallest quant17.2 GB
Quick answer

Rys Saber needs about 27.7 GB of memory at Q4_K_M (a 21.7 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 5 quants, from 17.2 GB (Q3_K_M) to 37.8 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 22.0 GB.

Rys Saber – a 35.5B Qwen3.5 MoE model with SABER refusal ablation

Rys Saber is a 35.5B‑parameter Qwen3.5 MoE model from the Ornstein family. It uses the SABER technique to remove refusal behavior while keeping full capability. The model requires a patched llama.cpp that supports per‑layer metadata.

Good at

  • Uncensored reasoning output
  • Large context handling
  • MoE efficiency

Limits

  • Requires patched llama.cpp
  • No vision support
  • Large VRAM needed for higher quants

Best for

  • Chatting with detailed reasoning
  • Long‑form content generation
  • Research prototyping

Who it’s for. Developers and researchers who need a large Qwen3.5 MoE model with uncensored reasoning and can run a patched llama.cpp. It is suitable for advanced LLM experimentation and custom deployment.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q3_K_M17.2 GB 23.2 GB
Too big Get file →
Q4_K_M21.7 GB 27.7 GB
Too big Get file →
Q5_K_M25.3 GB 31.4 GB
Too big Get file →
Q6_K29.2 GB 35.2 GB
Too big Get file →
Q8_037.8 GB 43.8 GB
Too big Get file →

For 8 GB VRAM, use Q3_K_M (17.17 GB file, ~23.2 GB memory at 8k context). For 12 GB VRAM, Q4_K_M (21.72 GB file, ~27.7 GB memory). For 16 GB VRAM, Q5_K_M (25.35 GB file, ~31.4 GB memory). Higher quants need more VRAM.

What GPU do you need for Rys Saber?

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: the best quant that fits is Q3_K_M (17.2 GB file, about 23.2 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (25.3 GB file, about 31.4 GB in use).
  • 48 GB VRAM: the best quant that fits is Q8_0 (37.8 GB file, about 43.8 GB in use).

Which Rys Saber quant should you choose?

Q4_K_M is the usual starting point for Rys Saber: it balances file size and quality well. Here it is a 21.7 GB file.

The Q8_0 file is 37.8 GB, about 1.7 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 Q3_K_M at 17.2 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35moe
Parameters35.5B (approximate, mixture of experts)
Active parameters3B
Layers41
Trained context262,144 tokens
KV cache (fp16)656 KB per token

Source

Quantized byGestaltLabs
FamilyOrnstein
Licenseapache-2.0
Downloads (HF)1,127
Last updated2026-04-18

Frequently asked questions about Rys Saber

How much VRAM does Rys Saber need?

About 27.7 GB at Q4_K_M with an 8K context window. That is the 21.7 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 Rys Saber on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 23.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Rys Saber GGUF?

It depends on the quant: Q3_K_M is 17.2 GB, Q4_K_M is 21.7 GB, Q5_K_M is 25.3 GB, Q6_K is 29.2 GB, Q8_0 is 37.8 GB.

What context length does Rys Saber support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 5.5 GB at 8K and 22.0 GB at 32K.

Where can I download Rys Saber GGUF?

The files are on Hugging Face in the GestaltLabs/Ornstein3.6-35B-A3B-RYS-SABER-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 Rys Saber 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 Rys Saber on 8 GB VRAM?

Use the Q3_K_M quant, which requires about 23.2 GB of VRAM at 8k context, so 8 GB is insufficient.

Do I need a special llama.cpp build?

Yes, a patched llama.cpp that reads per‑layer metadata is required to load this model.

Is the model censored?

No, it is a fully uncensored version with SABER refusal ablation.

Similar models

Browse all →

More from GestaltLabs

All by GestaltLabs →

More Ornstein models

All Ornstein →

Other large models

All large models →