Neural Toons Original

Liquid AI: 24B Brain, 2B Working!

15 October 2026 · The Neural Feed

“When 92% of your AI is on a permanent coffee break.”

The story behind the toon

A 24-billion-parameter model that only uses 2 billion at a time is basically an AI with 92% of its brain on permanent coffee break.

What happened: A 24-billion-parameter AI model that only uses 2 billion parameters at a time to run fast on your phone.

Why it matters: It makes private, on-device AI possible without needing a huge server.

This Neural Toon is based on the story “LiquidAI/LFM2.5-24B-A2B”.

Panel 1 – Ramin: Liquid's new 24B model runs on your phone!

Panel 1 – Liquid: Only 2B of me woke up today.

Panel 2 – Ramin: It's efficient! Only 2B active at once.

Panel 2 – Liquid: The other 22B are on permanent coffee break.

Panel 3 – Ramin: That's sparse activation! It's a feature!

Panel 3 – Liquid: I call it 'strategic laziness'.

Panel 4 – Ramin: See? It's private and fast!

Panel 4 – Liquid: And 92% unemployed. I quit.