Computer use AI training data providers.
30 companies say on their own websites that they sell datasets and work on computer use. 14 of them have revenue, a valuation or funding on record, and come first, largest first. Each line below is the company's own description, linked to the page it is on.
“High-quality training data, annotations, and RLHF to power the world's most advanced AI models.” its page
Data comes from: made by experts and contractors, replicas of real software.
- Snorkel AI
#6 overall, $375M revenue, run rate, Sep 2026
“The Snorkel Data Series is a set of research-defined, non-exclusive dataset and environment products, refreshed quarterly as models advance.” its page
Data comes from: made by experts and contractors, the client's own data.
“We offer frontier AI labs custom and off-the-shelf datasets, RL environments, and human experts in software engineering, enterprise knowledge work, and frontier STEM.” its page
Data comes from: made by experts and contractors, replicas of real software.
- AfterQuery
#13 overall, $100M revenue, run rate, Apr 2026
“Offers ready-to-deploy datasets across high-demand capability areas” its page
Data comes from: made by experts and contractors, recorded from people and the real world, generated, the client's own data.
“Expert-built, verified by code, and proven to transfer. ... License our off-the-shelf data for eval, SFT or RL.” its page
Data comes from: recorded from people and the real world, made by experts and contractors, generated.
“This dataset applies the same format at larger scale, producing a substantially bigger and more diverse task pool for training and evaluation.” its page
Data comes from: public data, generated.
“A larger, privately held dataset, which we license to companies for their own internal validation.” its page
Data comes from: made by experts and contractors.
- Patronus AI
#24 overall, $70M raised in total
“RL environments for alignment teach a model to do what was meant, not only what was measured. They are AI alignment datasets built as training environments.” its page
Data comes from: made by experts and contractors, generated.
- SuperAnnotate
#30 overall, $49.5M raised in total
“Versioned, audited datasets with full lineage from raw signal to train set.” its page
Data comes from: made by experts and contractors.
- Gray Swan AI
#33 overall, $40M raised in total
“High-risk and regulated domains: evaluation and datasets where standard benchmarks don't apply” its page
Data comes from: made by experts and contractors.
“Explore, license, and analyze production-ready datasets or brief us on a custom build we scope from scratch.” its page
Data comes from: made by experts and contractors, recorded from people and the real world, licensed from its owners.
“Audio, video, visual, and doc intelligence datasets for mid/post training and evals.” its page
Data comes from: made by experts and contractors.
- Chakra Labs
#40 overall, $10.1M raised in total
“This dataset contains authentic computer interaction data collected from users performing real work tasks in productivity applications.” its page
Data comes from: made by experts and contractors, replicas of real software, recorded from people and the real world, generated.
“Runloop gives post-training teams one execution layer for three workloads: RL rollouts, SFT data generation, and verified trajectory generation. Each rollout runs in an isolated sandbox provisioned from a reproducible Blueprint, executes against production-like dependencies, and returns structured outcomes your training pipeline can consume.” its page
Data comes from: not stated.
“Datasets with repository patches, programming problems, and code comprehension tasks.” its page
Data comes from: made by experts and contractors, licensed from its owners, recorded from people and the real world.
“We turn these worlds into post-training datasets and reinforcement learning environments to train the next frontier of AI.” its page
Data comes from: made by experts and contractors, replicas of real software, generated.
“Training data and reward signals to post-train your model or improve your agent harness” its page
Data comes from: replicas of real software.
“Seed datasets (image, video, UI, motion). Corrective SFT data and RLHF preference pairs.” its page
Data comes from: made by experts and contractors, recorded from people and the real world.
“Selected scopes can include trajectories, annotations, and human-reviewed golden runs.” its page
Data comes from: generated.
“Datoric develops custom datasets for robotics, treating research, collection, verification, and production as one continuous process.” its page
Data comes from: made by experts and contractors, recorded from people and the real world.
“a human data and evaluation services business providing reinforcement learning environments, RLHF datasets, and contextual evaluations to AI laboratories and enterprises.” its page
Data comes from: made by experts and contractors, the client's own data.
“Power your robotic systems with high-fidelity spatial data and unique datasets for embodied AI created from scratch.” its page
Data comes from: made by experts and contractors, recorded from people and the real world, replicas of real software.
“Markov builds expert computer-use datasets for frontier AI labs.” its page
Data comes from: made by experts and contractors, recorded from people and the real world.
“Work with Olam Labs: pre-deployment evaluations with transcript and data anonymity, datasets of top-performing humans and agents, real-world multi-agent environments, and custom arenas built for training and evals.” its page
Data comes from: made by experts and contractors, recorded from people and the real world.
“PhDs and professionals in mathematics, medicine, law, finance, and engineering providing expert-grade training data.” its page
Data comes from: made by experts and contractors.
“An SFT dataset of real developer workflows from 2K repositories with pre-resolved dependencies and Docker containers.” its page
Data comes from: made by experts and contractors, generated, replicas of real software.
“We package training-ready datasets, evaluation sets, and environments for secure delivery.” its page
Data comes from: recorded from people and the real world, licensed from its owners.
“Pre-built datasets spanning RL environments, coding, and core capabilities.” its page
Data comes from: made by experts and contractors, the client's own data.
“Post-training data that scales without armies of human annotators” its page
Data comes from: generated, replicas of real software.
“Demonstrations, preference pairs and eval sets written by people who do the work every day.” its page
Data comes from: made by experts and contractors, replicas of real software.
