With Mobius Labs' Aana models, we're bringing deeper multimodal understanding to Dropbox Dash

Dropbox has acquired AI startup Mobius Labs and is integrating their multimodal AI models, called Aana, into Dropbox Dash to enable deeper understanding of rich media content.

• •Aana combines open-source foundation models for speech, vision, and language using transformer-based and mixture-of-experts (MoE) architectures optimized for off-the-shelf GPUs
• •The system processes audio, video, and images together rather than as separate streams, mapping everything into a shared vector space for fast multimodal search
• •HQQ enables low-bit (8-bit and 4-bit) inference to dramatically reduce compute and memory costs, while Gemlite accelerates matrix multiplications and attention layers with custom GPU kernels
• •The Aana SDK handles batching, model coordination, and GPU utilization, allowing teams to configure and deploy multimodal pipelines with minimal overhead
• •The goal is to enable agentic workflows that can analyze multimedia data, surface insights automatically, and act on behalf of team