unsloth
Qwen3.8 27B
Qwen3.8 27B is a 27.3B dense qwen35 model from unsloth. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters27.3B
Context256K
Attentionhybrid
Licenseapache-2.0
64 layers · 4 KV heads · head dim 256 · 48 linear-attention layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 50.9 GB | Hugging Face file bytes |
| UD-Q8_K_L | 26.1 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 29.3 GB | Hugging Face file bytes |
| Q8_0 | 27.1 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 23.6 GB | Hugging Face file bytes |
| UD-Q6_K_L | 22.5 GB | Hugging Face file bytes |
| UD-Q6_K_M | 21.5 GB | Hugging Face file bytes |
| UD-Q6_K | 20.5 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 19.4 GB | Hugging Face file bytes |
| UD-Q5_K_M | 18.4 GB | Hugging Face file bytes |
| UD-Q5_K_S | 17.4 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 16.4 GB | Hugging Face file bytes |
| UD-Q4_K_M | 15.3 GB | Hugging Face file bytes |
| UD-Q4_K_S | 14.3 GB | Hugging Face file bytes |
| UD-IQ4_XS | 13.3 GB | Hugging Face file bytes |
| Q4_1 | 16.3 GB | Hugging Face file bytes |
| Q4_0 | 15.0 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 12.2 GB | Hugging Face file bytes |
| UD-IQ3_S | 11.2 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 10.2 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 9.15 GB | Hugging Face file bytes |
| UD-IQ2_S | 7.80 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 6.77 GB | Hugging Face file bytes |
| UD-IQ1_M | 6.27 GB | Hugging Face file bytes |
| UD-IQ1_S | 5.77 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | UD-IQ1_M | 23–31 |
| GeForce RTX 3060 12GB | Tight fit | UD-Q2_K_XL | 19–26 |
| GeForce RTX 4060 Ti 16GB | Tight fit | UD-IQ4_XS | 11–15 |
| GeForce RTX 4090 | Tight fit | UD-Q6_K | 25–34 |
| GeForce RTX 5090 | Tight fit | UD-Q8_K_L | 34–47 |
| Apple M4 Pro | Tight fit | UD-Q4_K_XL | 8.3–11 |
| Apple M5 | Tight fit | UD-Q2_K_XL | 8.1–11 |
| CPU only · DDR5-5600 | Runs on CPU | UD-Q6_K_XL | 1.9–2.6 |
Smallest devices that fit it fully
- GeForce RTX 5060 Ti 8GB — Tight fit, UD-IQ1_M, 35–48 tok/s
- GeForce RTX 5060 — Tight fit, UD-IQ1_M, 35–48 tok/s
- GeForce RTX 5050 — Tight fit, UD-IQ1_M, 25–34 tok/s
- GeForce RTX 4060 Ti 8GB — Tight fit, UD-IQ1_M, 23–31 tok/s
- GeForce RTX 4060 — Tight fit, UD-IQ1_M, 21–29 tok/s
- GeForce RTX 3070 Ti — Tight fit, UD-IQ1_M, 48–65 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-IQ4_XS).
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-IQ4_XS -c 8192 -ngl 64
Ollama
ollama run qwen3.8
LM Studio
lms get unsloth/Qwen3.8-27B-GGUF # Then load quant UD-IQ4_XS