Mistral AI · Apache 2.0
Mistral Small 4
A 119B mixture-of-experts model that merges Mistral’s chat, reasoning and coding lines. For 96–128 GB machines.
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 | At 128K |
|---|---|---|---|---|---|
| 119B (6.5B active) | Q4_K_M | 72.6 GB | 74.7 GB | 75.3 GB | 77.5 GB |
| Q8_0 | 126 GB | 129 GB | 129 GB | 131 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 | 119B Q4_K_M | Too big | — |
| RTX 4060 · 8 GB | 119B Q4_K_M | Too big | — |
| RTX 3060 · 12 GB | 119B Q4_K_M | Too big | — |
| RTX 5060 Ti · 16 GB | 119B Q4_K_M | Too big | — |
| RTX 4090 · 24 GB | 119B Q4_K_M | Too big | — |
| RTX 5090 · 32 GB | 119B Q4_K_M | Too big | — |
| Mac with M4 · 16 GB | 119B Q4_K_M | Too big | — |
| Mac with M4 Pro · 48 GB | 119B Q4_K_M | Too big | — |
| Mac with M5 Max · 128 GB | 119B Q4_K_M | Runs great | 45–95tok/sFast |
| Ryzen AI Max+ 395 · 128 GB | 119B Q4_K_M | Runs well | 20–45tok/sSmooth |
| DGX Spark · 128 GB | 119B Q4_K_M | Runs great | 25–45tok/sSmooth |
| RTX PRO 6000 · 96 GB | 119B Q4_K_M | Runs well | 100–200tok/sFast |
| Mac Studio with M5 Ultra · 256 GB | 119B Q8_0 | Runs great | 30–70tok/sFast |
| Mac Studio with M3 Ultra · 512 GB | 119B Q8_0 | Runs great | 25–50tok/sSmooth |
What it is good and bad at
Good at
- Fast for its size: 6.5B parameters work per word
- Reads images
Not good at
- Q4 file is 73 GB
- No Ollama library entry yet