unsloth
SmolLM3 3B
SmolLM3 3B is a 3.08B dense smollm3 model from unsloth. Configured context length is 65,536 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters3.1B
Context64K
Attentiongqa
Licenseapache-2.0
36 layers · 4 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 5.74 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 3.63 GB | Hugging Face file bytes |
| Q8_0 | 3.05 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 2.57 GB | Hugging Face file bytes |
| Q6_K | 2.36 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 2.06 GB | Hugging Face file bytes |
| Q5_K_M | 2.06 GB | Hugging Face file bytes |
| Q5_K_S | 2.01 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 1.81 GB | Hugging Face file bytes |
| Q4_K_M | 1.78 GB | Hugging Face file bytes |
| Q4_K_S | 1.69 GB | Hugging Face file bytes |
| IQ4_NL | 1.69 GB | Hugging Face file bytes |
| IQ4_XS | 1.61 GB | Hugging Face file bytes |
| Q4_1 | 1.85 GB | Hugging Face file bytes |
| Q4_0 | 1.69 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 1.50 GB | Hugging Face file bytes |
| Q3_K_M | 1.46 GB | Hugging Face file bytes |
| Q3_K_S | 1.33 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 1.20 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 1.20 GB | Hugging Face file bytes |
| Q2_K | 2.33 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 0.89 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | BF16 | 23–32 |
| GeForce RTX 3060 12GB | Runs | BF16 | 29–40 |
| GeForce RTX 4060 Ti 16GB | Runs great | BF16 | 23–32 |
| GeForce RTX 4090 | Runs great | BF16 | 82–112 |
| GeForce RTX 5090 | Runs great | BF16 | 146–199 |
| Apple M4 Pro | Runs great | BF16 | 22–30 |
| Apple M5 | Runs | BF16 | 12–17 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 7.3–9.9 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Tight fit, UD-Q8_K_XL, 21–28 tok/s
- GeForce RTX 2060 6GB — Tight fit, UD-Q8_K_XL, 41–56 tok/s
- GeForce GTX 1660 Super — Tight fit, UD-Q8_K_XL, 41–56 tok/s
- GeForce GTX 1660 Ti — Tight fit, UD-Q8_K_XL, 35–48 tok/s
- Intel Arc A380 — Tight fit, UD-Q8_K_XL, 23–31 tok/s
- GeForce RTX 5060 Ti 8GB — Tight fit, BF16, 36–50 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (BF16).
llama.cpp
llama-server -hf unsloth/SmolLM3-3B-GGUF:BF16 -c 8192 -ngl 36
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/SmolLM3-3B-GGUF # Then load quant BF16