REQUIREMENTS
System Requirements — BOM AI
BOM AI runs on your computer. The more powerful the machine, the faster and smarter it gets
Minimum for Every Model
| Item | Minimum |
|---|---|
| Operating system | Windows 10/11 64-bit · macOS 12+ · Ubuntu 22.04+ |
| RAM | 8 GB |
| Disk space | 10 GB free (including models) |
| Internet | Only for the first download |
| GPU | Optional, but highly recommended |
Recommended GPU per Model
🟢 Light — 8 GB RAM (or no GPU)
Model: BOM AI Light (Qwen3.5 4B) (~3.4 GB) · Thai chat + image reading + translation
Example cards: NVIDIA GTX 1660 · RTX 3050 · RTX 4060
Best for: everyday laptops — Thai chat + translation
No GPU at all? It still works — the CPU is slower but every feature is included
Why a Graphics Card?
AI models work by computing billions of matrix operations. GPUs (graphics cards) are designed exactly for this kind of math — 10-50x faster than CPUs
The smarter the model, the more parameters it has, and the more VRAM (memory on the graphics card) it needs. That is why graphics cards are expensive — they genuinely matter for AI
| Model size | Parameters | VRAM used | Response speed |
|---|---|---|---|
| 4B | 4 billion | ~2.7 GB | Very fast |
You choose the edition — both editions
BOM AI ships 2 editions and you pick the one you want. The app detects your GPU at startup and recommends a good fit, but the choice is yours.
How to Check Your GPU
Windows: Press Ctrl+Shift+Esc → Performance tab → GPU to see the model name and VRAM
macOS: Apple menu → About This Mac → check Graphics
Linux: Open Terminal and type nvidia-smi (if you have NVIDIA) or lspci | grep VGA
Locked Generation Configuration
BOM AI ships with a locked configuration (temperature 0.6, top_p 0.95, top_k 20, repeat_penalty 1.0, num_ctx 8192, num_predict 4096). These settings are optimized for quality and safety. They represent maximum context and generation limits — actual usage adapts to the task at hand.
