Hugging Face · Apache 2.0
SmolLM3 3B
A fully open 3B model from Hugging Face: weights, data and training recipe all published.
How much memory it needs
The model file, plus memory for the conversation, plus about 1 GB for the app. Longer conversations need more.
| Size | Version | File | At 8K | At 32K |
|---|---|---|---|---|
| 3.1B | Q4_K_M | 1.9 GB | 3.5 GB | 5.3 GB |
| Q8_0 | 3.3 GB | 4.9 GB | 6.7 GB |
Which computers run it
The best size for each machine at an 8K-token conversation. Speeds are estimates from memory bandwidth.
| Computer | Best version | Fit | Speed |
|---|---|---|---|
| Laptop or PC, no graphics card, 32 GB | 3.1B Q8_0 | Runs great | 14–20tok/sSmooth |
| RTX 4060 · 8 GB | 3.1B Q8_0 | Runs great | 40–65tok/sFast |
| RTX 3060 · 12 GB | 3.1B Q8_0 | Runs great | 55–85tok/sFast |
| RTX 5060 Ti · 16 GB | 3.1B Q8_0 | Runs great | 65–100tok/sFast |
| RTX 4090 · 24 GB | 3.1B Q8_0 | Runs great | 140–210tok/sFast |
| RTX 5090 · 32 GB | 3.1B Q8_0 | Runs great | 220–340tok/sFast |
| Mac with M4 · 16 GB | 3.1B Q8_0 | Runs great | 18–30tok/sSmooth |
| Mac with M4 Pro · 48 GB | 3.1B Q8_0 | Runs great | 40–65tok/sFast |
| Mac with M5 Max · 128 GB | 3.1B Q8_0 | Runs great | 80–140tok/sFast |
| Ryzen AI Max+ 395 · 128 GB | 3.1B Q8_0 | Runs great | 40–60tok/sFast |
| DGX Spark · 128 GB | 3.1B Q8_0 | Runs great | 40–65tok/sFast |
| RTX PRO 6000 · 96 GB | 3.1B Q8_0 | Runs great | 220–340tok/sFast |
| Mac Studio with M5 Ultra · 256 GB | 3.1B Q8_0 | Runs great | 120–200tok/sFast |
| Mac Studio with M3 Ultra · 512 GB | 3.1B Q8_0 | Runs great | 85–140tok/sFast |
What it is good and bad at
Good at
- Fully open
- Thinking mode
Not good at
- 3B model: limited knowledge
- No Ollama library entry