unsloth
Llama 3.2 1B Instruct
Llama 3.2 1B Instruct is a 1.24B dense llama model from unsloth. Configured context length is 131,072 tokens. License: llama3.2. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters1.2B
Context128K
Attentiongqa
Licensellama3.2
16 layers · 8 KV heads · head dim 64 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 2.31 GB | Hugging Face file bytes |
| F16 | 2.31 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 1.72 GB | Hugging Face file bytes |
| Q8_0 | 1.23 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 1.11 GB | Hugging Face file bytes |
| Q6_K | 0.95 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 0.86 GB | Hugging Face file bytes |
| Q5_K_M | 0.85 GB | Hugging Face file bytes |
| Q5_K_S | 0.83 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 0.78 GB | Hugging Face file bytes |
| Q4_K_M | 0.75 GB | Hugging Face file bytes |
| Q4_K_S | 0.72 GB | Hugging Face file bytes |
| IQ4_NL | 0.72 GB | Hugging Face file bytes |
| IQ4_XS | 0.69 GB | Hugging Face file bytes |
| Q4_1 | 0.77 GB | Hugging Face file bytes |
| Q4_0 | 0.72 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 0.67 GB | Hugging Face file bytes |
| Q3_K_M | 0.64 GB | Hugging Face file bytes |
| Q3_K_S | 0.60 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 0.54 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 0.56 GB | Hugging Face file bytes |
| Q2_K | 1.08 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 0.43 GB | Hugging Face file bytes |
| UD-IQ1_M | 0.41 GB | Hugging Face file bytes |
| UD-IQ1_S | 0.39 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Runs great | BF16 | 58–79 |
| GeForce RTX 3060 12GB | Runs great | BF16 | 72–98 |
| GeForce RTX 4060 Ti 16GB | Runs great | BF16 | 58–79 |
| GeForce RTX 4090 | Runs great | BF16 | 202–275 |
| GeForce RTX 5090 | Runs great | BF16 | 359–489 |
| Apple M4 Pro | Runs great | BF16 | 55–75 |
| Apple M5 | Runs great | BF16 | 31–42 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 18–24 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Runs great, BF16, 34–46 tok/s
- GeForce RTX 2060 6GB — Runs great, BF16, 67–92 tok/s
- GeForce GTX 1660 Super — Runs great, BF16, 67–92 tok/s
- GeForce GTX 1660 Ti — Runs great, BF16, 58–79 tok/s
- Intel Arc A380 — Runs great, BF16, 37–51 tok/s
- GeForce RTX 5060 Ti 8GB — Runs great, BF16, 90–122 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (BF16).
llama.cpp
llama-server -hf unsloth/Llama-3.2-1B-Instruct-GGUF:BF16 -c 8192 -ngl 16
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/Llama-3.2-1B-Instruct-GGUF # Then load quant BF16