MosaicMind raises $45M seed to commercialize on-device LLM inference

AI · 3 min read

MosaicMind, which builds compact LLM runtimes for mobile and embedded devices, announced a $45 million seed round led by NextWave Ventures with contributions from AI-native funds. The money will be used to expand engineering around quantized models, compiler optimizations, and cross-platform SDKs.

The company’s tech combines model distillation with hardware-aware compilation to run 7B-parameter class models on modern mobile NPUs and compute accelerators. MosaicMind highlights offline capabilities—local summarization, private code completions, and on-device personalization—helpful for apps in healthcare, finance, and enterprise where data residency matters.

MosaicMind also unveiled partnerships with two OEMs for trial integration on mid-range devices slated for late 2026. The startup plans to offer a hosted tuning service for enterprises that want tight on-device model adaptation while keeping source data local to the device.