Datoric licenses robotics and physical-AI training data (egocentric video, manipulation and teleoperation recordings), runs custom collections to a buyer's specification, and also offers computer-use, voice and gameplay datasets.
What they offerTraining data, Human data
Type of workComputer use, Browser, Enterprise workflows, Robotics and physical, Games, Multi-domain, Other
Good Start Labs builds game-based reinforcement-learning environments, turns gameplay into human and synthetic training data, and publishes evaluation benchmarks for AI labs. It also helps game publishers capture gameplay data and put AI agents into their games.
What they offerRL environments, Human data, Evals and benchmarks, Training data
Type of workGames
Data sourcesBuilt from customer data, Licensed real-world data
Intelligence runs DesignArena, where people vote between AI-generated websites, games, images, video and audio, and sells model evaluations to AI labs. It says it shares users' prompts, outputs and votes with AI model developers, who may train on them, after taking steps to remove identifiers like usernames; its terms let it charge for that use.
What they offerEvals and benchmarks, Human data
Type of workCoding, Creative, Games, Enterprise workflows
Markov sells computer-use datasets and task environments—including expert CAD, productivity, coding, and gaming workflows—to AI labs and developers training or evaluating desktop agents.
What they offerRL environments, Training data, Evals and benchmarks, Human data
Type of workMulti-domain, Computer use, Coding, Browser, Games
Midcentury sells real-world human-action and gameplay datasets, along with cloud simulation tools, to robotics developers, AI labs, and enterprises training and evaluating physical AI systems.
What they offerTraining data, RL environments, Simulation, Human data
Olam Labs builds simulated multi-agent environments, behavioral evaluations and game-derived datasets for AI labs and researchers training or evaluating models’ social, agentic and safety capabilities.
What they offerRL environments, Training data, Evals and benchmarks, Simulation
Type of workCoding, Computer use, Enterprise workflows, Finance, Games, Security, Multi-domain
Prime Intellect sells AI labs and developers an open stack for reinforcement-learning environments, model evaluations, hosted training, inference, sandboxes, and GPU compute to build and improve agentic models.
What they offerRL environments, Evals and benchmarks, RL infrastructure
Quesma publishes coding-agent benchmarks and builds RL training environments for frontier labs; it also markets tools that help development teams understand agent costs, workflows and outcomes.
What they offerRL environments, Evals and benchmarks
Sieve supplies AI labs and model developers with curated or custom-collected video, audio, image, and interaction data, plus annotations, evaluation sets, and interactive environments.
What they offerTraining data, Evals and benchmarks, RL environments, Human data
Type of workCreative, Computer use, Robotics and physical, Games, Multi-domain
Data sourcesLicensed real-world data, Expert or contractor-created
Last checked
Initial directory snapshot · exported Nov 4, 2026. Original source check dates are retained. Automatic monitoring is not active yet. How we check information