A whole AI brain fits in about one app of space β and it keeps thinking when your signal bars vanish.
Liquid AIβs LFM3 is free, open, and runs right on your phone or laptop. It sees photos, hears you, talks back, and can learn a job you give it.
Pick a door β all free:
| Door | What you get |
|---|---|
| Omni in a browser tab β sees, hears, speaks, no install | |
ollama run lfm3, then switch off the wifi and keep chatting |
|
| CPU-only browser build β no WebGPU needed |
Your mischief list
-
Make it your own expert on Googleβs free GPU β a few hours, one job -
Grab the 1.05 GB smart 4-bit file β trained to keep its smarts at tiny size -
Talk over it and cut in β a voice agent that lives in your browser -
Let it decide instead of chat β many questions answered at once, with confidence scores -
Try the community remix β same brain, 1.58Γ faster on one Mac
ποΈ More free doors
A 262K-memory brain by API β $0 in and out, no download
A 45-line offline chat app β no server, account or key
Run the vision model on a free GPU Β· a coding workspace in a tab
A phone app that does the chores β notes, to-dos, calendar, alarms, camera questions, all on-device
The full 256K memory in ~4.3 GB Β· all four models under 5 GB
Voice + webcam agent Β· a Pi making image decisions
π The score, straight from the maker
- 71.4 β the vision modelβs average over 62 vision tests with thinking on, against 70.0 for Gemini 3.5 Flash Lite
- 62.1 β the same model with thinking off; Gemma 4 12B (71.7) and GPT-5.6 Luna (73.4) still lead the table
- ~1 GB is the weight file; running memory on an iPhone peaked at 1.9 GB with a 32K prompt
- Every figure above is Liquidβs own; the announcement and the scoreboard written up carry them
- Free commercial use runs to $10M a year in revenue; the weights are open
Go on β switch off the wifi.
!