Labelbox, founded in 2018 and based in San Francisco, builds infrastructure for creating and refining the large-scale training datasets frontier AI models depend on, spanning data curation, human- and model-assisted labeling, and model evaluation across three connected products: Catalog, Annotate, and Model.
That full-pipeline scope, curate data, label it, then evaluate the resulting model, positions Labelbox as a reinforcement-learning data factory rather than a single-purpose labeling vendor, a distinction that has drawn AI labs and enterprises building frontier models and production AI applications, backed by $189 million in total funding.
Pricing runs on Labelbox Units (LBUs), a normalized unit of data work priced at $0.10 per LBU starting on the Starter tier, with a capped free tier for evaluation and custom Enterprise pricing above that; storing and curating data consumes LBUs at low cost, labeling costs more, and model evaluation sits in between. For an AI lab or enterprise that needs the full data pipeline, curation through evaluation, not only a labeling queue, Labelbox's connected product suite addresses that directly.
