tip/26
Match model to RAM.
A rough rule for local models: a 4-bit 7–8B model needs about 5–6GB of memory. Start small, then step up if it's too weak.
Ollama · Local & Open Models26 / 32
teechmeai/tips30-second
30-second
wins.save these!
Small habits that make AI noticeably better. Read one, use it today. Prompts are highlighted, so copy and go.
32 tips · 8 topics · 0 fluff
start here ↓
tip/27try it today
Keep the model warm.
Ollama unloads idle models after a few minutes. Set keep_alive on requests if you want instant replies during a session.
Ollama · Local & Open Models27 / 32