Cactus Compute

Cactus Compute (YC S25) is an on-device AI inference company providing a unified SDK and runtime engine that enables developers to run large language models, speech transcription, and vision models directly on smartphones, laptops, and wearable edge devices. The open-source Cactus Engine delivers sub-120ms on-device latency with hardware-specific acceleration and zero-copy memory mapping, with a hybrid cloud fallback that routes only complex requests to the cloud to achieve 5x cost savings versus cloud-only inference. The platform supports iOS, Android, macOS, and wearables from a single cross-platform codebase.