Voxel51 has released Voxel51 3.0, an end-to-end multimodal data platform that integrates search, curation, annotation, and evaluation into a single system designed for teams training models on sensor data from robots, vehicles, and factories. The major update expands the platform beyond visual data by introducing time-series multimodal support, unified annotation for 2D, 3D, video, and multimodal inputs, and Agentic Labeling driven by vision-language models. Additionally, the release introduces the Voxel51 Agent with over 45 Model Context Protocol tools, a comprehensive data provenance layer with immutable dataset versioning, and free self-service access tiers.
The platform utilizes a persistent hybrid index powered by LanceDB to deliver semantic, lexical, and structured retrieval capabilities, allowing teams to find specific moments across multiple sensor streams on a single timeline. Voxel51 3.0 remains fully model-agnostic, enabling organizations to use foundation models for bootstrapping before swapping in proprietary domain models for embeddings, auto-labeling, inference, and evaluation. Workflows are governed by activity logs and dataset-level provenance tracking, ensuring reproducibility across training runs.
This release streamlines complex data workflows for physical AI development by reducing time spent on manual curation and labeling.