LiquidAI d1-2B Runs Locally: 14 Decisions Per Second With Zero Tokens

Developer Ivan Fioravanti ran LiquidAI's d1-2B decision model locally on Apple Silicon M5 Max via mlx-vlm, reading camera gestures and making about 14 decisions per second with 68ms latency and zero token generation.

2026-11-16 ~ 2026-11-17 · 2 related posts