European AI lab Desert Ant launched 18 small on-device models for audio, vision, and text. No round-trips, no tokens, no data leaving the device.
- Desert Ant Labs, a European AI startup, has launched 18 small on-device models for audio, vision, and text tasks, available via a single SDK for Swift, Kotlin, and JavaScript.
- The models are designed to replace cloud API calls for repetitive tasks: a 9MB audio model runs 302x faster than real time on an iPhone 16 Pro, and a 12MB PII redaction model catches 88.8% of personal data versus a 2.3GB competitor catching 91.1%.
- All models are free up to 100k monthly active devices, no tokens or logins required, which changes the cost calculus for developers currently paying per call for tasks that run identically thousands of times a day.
