unsloth
Llama 3.1 8B Instruct
Llama 3.1 8B Instruct is a 8.03B dense llama model from unsloth. Configured context length is 131,072 tokens. License: llama3.1. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters8.0B
Context128K
Attentiongqa
Licensellama3.1
32 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 15.0 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 9.85 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 6.83 GB | Hugging Face file bytes |
| Q6_K | 6.14 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 5.34 GB | Hugging Face file bytes |
| Q5_K_M | 5.34 GB | Hugging Face file bytes |
| Q5_K_S | 5.21 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 4.65 GB | Hugging Face file bytes |
| Q4_K_M | 4.58 GB | Hugging Face file bytes |
| Q4_K_S | 4.37 GB | Hugging Face file bytes |
| IQ4_NL | 4.36 GB | Hugging Face file bytes |
| IQ4_XS | 4.16 GB | Hugging Face file bytes |
| Q4_1 | 4.78 GB | Hugging Face file bytes |
| Q4_0 | 4.35 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 3.91 GB | Hugging Face file bytes |
| Q3_K_M | 3.74 GB | Hugging Face file bytes |
| Q3_K_S | 3.41 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 3.09 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 3.16 GB | Hugging Face file bytes |
| Q2_K | 6.04 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 2.33 GB | Hugging Face file bytes |
| UD-IQ1_M | 2.13 GB | Hugging Face file bytes |
| UD-IQ1_S | 2.02 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | UD-Q5_K_XL | 23–32 |
| GeForce RTX 3060 12GB | Runs | UD-Q6_K_XL | 24–32 |
| GeForce RTX 4060 Ti 16GB | Runs | UD-Q8_K_XL | 14–19 |
| GeForce RTX 4090 | Runs | BF16 | 32–44 |
| GeForce RTX 5090 | Runs great | BF16 | 57–78 |
| Apple M4 Pro | Tight fit | BF16 | 8.8–12 |
| Apple M5 | Tight fit | UD-Q8_K_XL | 7.6–10 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 2.9–3.9 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Tight fit, IQ4_XS, 18–25 tok/s
- GeForce RTX 2060 6GB — Tight fit, IQ4_XS, 37–50 tok/s
- GeForce GTX 1660 Super — Tight fit, IQ4_XS, 37–50 tok/s
- GeForce GTX 1660 Ti — Tight fit, IQ4_XS, 32–43 tok/s
- Intel Arc A380 — Tight fit, IQ4_XS, 20–28 tok/s
- GeForce RTX 5060 Ti 8GB — Tight fit, UD-Q5_K_XL, 36–49 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-Q8_K_XL).
llama.cpp
llama-server -hf unsloth/Llama-3.1-8B-Instruct-GGUF:UD-Q8_K_XL -c 8192 -ngl 32
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/Llama-3.1-8B-Instruct-GGUF # Then load quant UD-Q8_K_XL