I haven’t run this yet. It’s on my “next to try” list, and here’s why.
Why LFM3 caught my eye
Liquid AI’s LFM3, now in Ollama 0.40.12, is a 1.7B model aimed squarely at local agents: tool calling, structured output and a 256K-token context window. Small, local and built for agents is exactly what I like for automations that shouldn’t send data anywhere.
What I’ll check when I try it
- Platform support first. Ollama’s notes say it’s Apple-hardware-only in this release, with Linux and Windows “coming soon.”
- Does tool calling stay reliable over a long session, or does it drift?
- How does structured output hold up against bigger models on simple tasks?
- What does it try to reach on the network? Same drill as my phone-home checklist.
When I’ve actually put it through its paces, I’ll write it up.
God Bless, and May the Source Be with You!
