Desert Ant Labs: local, fast models that run on device

Desert Ant Labs is building small, specialized models that run locally on Apple, Android, and the web, with native SDKs aimed at embedding speech enhancement, PII redaction, structured extraction, moderation, vision, and other single-purpose capabilities directly into products.

The project’s published pitch is notably product-oriented rather than a general-purpose model: every model is free up to 100,000 monthly active devices per SDK, with no per-user invocation limit. That is a potentially useful deployment model for privacy, latency, and cloud-cost-sensitive applications, though the claims are vendor-provided and the HN discussion was the main external signal (340 points, 85 comments).