Moondream develops an open-source vision-language model with just 1.6 billion parameters that rivals much larger models. Its model (Moondream2) has logged over 2 million downloads and 5,100 GitHub stars and achieves 80.3% on VQAv2 and 64.3% on GQA. The company emphasizes energy efficiency and edge deployment—its CTO cited per-token consumption of about 0.6 joules per billion parameters—and the model can run on iOS and mobile phones. Moondream targets enterprise use cases where cloud costs and privacy are concerns, with applications in retail inventory, vehicle inspection, and air-gapped factory quality control. It has launched Moondream Cloud Service to provide a hybrid cloud-to-edge development path while avoiding vendor lock-in. With $4.5 million in fresh pre-seed funding, the company plans to expand its team and scale its technology while maintaining efficiency and accessibility.

Total raised

$5M

Funding rounds

1

Latest round

Seed

Latest activity

Oct 2024