EdgeMind Releases 'Whitaker' — A 1.5B-Parameter Offline LLM for Mobile Apps, Raises $18M

AI · 4 min read

Whitaker is a 1.5-billion-parameter transformer tuned for tasks like summarization, Q&A, and domain-specific assistants while running locally on high-end smartphones and edge devices. EdgeMind emphasizes that Whitaker supports sparse attention optimizations and low-memory retrieval augmentation, enabling effective multi-turn conversations without cloud dependency.

The release includes SDKs for Android, iOS, and embedded Linux, plus a model management portal for federated updates and telemetry. EdgeMind claims companies can maintain data residency and reduce inference costs by keeping models on-device, while still updating behavior via encrypted patch delivery.

The $18 million Series A will be used to build partnerships with handset makers, expand support to wearables, and fund research into quantization techniques that retain quality at lower bit-widths. Early adopters include a telemedicine app and a field-inspection vendor piloting offline guidance features.