OpenFrame unveils on-device multimodal LLM chip and raises $60M Series A

AI · 6 min read

The chip, codenamed Aurora, pairs specialized tensor cores for language inference with a vision accelerator tuned for high-resolution embeddings and cross-attention. OpenFrame demonstrated real-time multimodal apps—live captioning with scene-aware summaries—running solely on a compact development board.

The $60 million Series A was led by Foundry AI and includes strategic investments from a consumer electronics supplier planning to embed Aurora in next-year laptop and AR prototypes. OpenFrame will use the funds to refine silicon, expand SDKs, and establish manufacturing partnerships.

OpenFrame positions Aurora for privacy-sensitive use cases—healthcare notes, personal assistants, and offline field analytics—where constant cloud connectivity is either impossible or undesirable. Competitors say on-device chips are becoming mainstream, but OpenFrame's integration of multimodal accelerators sets it apart.