QW

Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid

QwenGGUF by webmp3Updated 2026-10-08
Download on Hugging Face
Parameters26.9B
Context256K
Layers64
Architectureqwen35
Smallest quant10.7 GB
Quick answer

Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid needs about 13.3 GB of memory at its main file (a 10.7 GB file) with an 8K context window, so it fits on a 16 GB GPU. Longer context needs more: at 32K tokens the KV cache alone adds about 8.6 GB.

Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid – experimental post‑quantized GGUF

Sakura is a 26.9B‑parameter Qwen model compressed to 10.67 GB. It is based on Qwen3.8‑27B‑EfficientThink and uses a custom GGUF runtime. The model supports reasoning and tools but is experimental.

Good at

  • Reasoning
  • Tools usage
  • Large context 262,144 tokens

Limits

  • Requires PrismML llama.cpp fork
  • No vision support
  • Perplexity regression

Best for

  • Local inference on high‑VRAM GPUs
  • Fine‑tuning experiments
  • Research prototyping

Who it’s for. Developers and researchers who need a compact 27B Qwen variant for local deployment and are willing to use a specialized llama.cpp fork. The model is suitable for those testing post‑quantization techniques.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—10.7 GB 13.3 GB
Too big Get file →

The only available quant is the 10.67 GB file, which needs about 13.3 GB of VRAM at 8 k context. For 8 GB VRAM, the model cannot run. For 16 GB VRAM, it is feasible with the given memory estimate. For 32 GB VRAM, it runs comfortably.

What GPU do you need for Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid?

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: the best quant that fits is the main file (10.7 GB file, about 13.3 GB in use).

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters26.9B
Layers64
Trained context262,144 tokens
KV cache (fp16)256 KB per token

Frequently asked questions about Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid

How much VRAM does Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid need?

About 13.3 GB at its main file with an 8K context window. That is the 10.7 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 Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 13.3 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid GGUF?

It depends on the quant: file is 10.7 GB.

What context length does Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.6 GB at 32K.

Where can I download Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid GGUF?

The files are on Hugging Face in the webmp3/Sakura-Qwen3.8-27B-EfficientThink-GSQ-RCO-Hybrid repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Sakura Qwen3.8 27B EfficientThink GSQ RCO Hybrid 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 Sakura on 8 GB VRAM?

No, the model requires approximately 13.3 GB of VRAM at 8 k context.

What runtime is required?

Sakura needs the PrismML llama.cpp fork with custom GGUF support; stock llama.cpp cannot load it.

Does Sakura support vision?

No, the model has no vision capabilities.

Similar models

Browse all →

More from webmp3

All by webmp3 →

More Qwen models

All Qwen →

Other medium models

All medium models →