Polaris Labs launches Hermes: an on-device multimodal LLM and raises $40M to commercialize edge AI

AI · 6 min read

Hermes runs with quantized weights and supports vision, audio, and text inputs for tasks like image-aware summarization and live captioning. Polaris built a toolkit to integrate Hermes into mobile apps with energy-aware scheduling and modular privacy rules.

The $40 million Series A was led by Evergreen Partners, who cited growing demand for edge-first AI that reduces cloud costs and improves latency. Polaris will invest proceeds in model optimization, developer tools, and partnerships with chipset makers for hardware acceleration.

Developers have early access to Hermes SDK and have already built prototypes like a camera-native note app and a real-time translation layer that combines on-device OCR with local translation. Polaris emphasizes that users retain control of data and can run the model entirely on-device for sensitive applications.