Skip to main content

Coop

Coop runs independent, industry-informed benchmarks on real robots, enabling model developers to compare checkpoints, understand failures, and measure progress with each release. Its first benchmark—a DO NOT ENTER sign-following test—ran four models across 220 trials on one humanoid robot, with results published in an open journal format. Coop also open-sourced Spur, a high-level motor controller for the Unitree R1.

Traction –/10
San Francisco, United States · HQ
10+ followers
  • Robotics
  • Data & Analytics
  • Artificial Intelligence
  • Developer Tools
  • Content & Publishing
Updated 10 days ago

Funding

Funding not disclosed

Funding rounds are not available yet.

Founders

Founder details are not available yet.

Product

Problem

Robot model developers and researchers lack trusted, comparable evaluation methods for physical systems. Simulation often fails to capture real-world complexity, and existing benchmarks become saturated or unreliable, leaving teams unable to determine whether a new model checkpoint is genuinely better than its predecessor.

Solution

Coop operates independent, industry-informed benchmarks on real robots, systematically comparing model checkpoints under matched starting conditions and documented resets. The company records successes, failures, and interventions across every trial, producing transparent data that reveals where models struggle and how they improve. Coop’s first benchmark—a DO NOT ENTER sign-following test—ran four models across 220 trials on one humanoid robot. Results are published in an open journal format alongside analysis and insights, helping researchers and developers make data-driven decisions about model releases.

Target Audience

Robot model developers, AI researchers, and industry teams who need rigorous physical evaluation data to compare checkpoints and guide model development decisions.

Features

  • Physical benchmark trials on real robots with matched starting conditions and documented resets
  • Standardized protocols that record successes, failures, and human interventions for each trial
  • Cross-model comparison of checkpoints to evaluate performance differences under identical conditions
  • Publication of raw trial data and playback footage for independent verification
  • A public research journal with field notes and technical analyses of benchmark outcomes
  • Open-source release of Spur, a high-level motor controller for the Unitree R1 that maintains balance and plans motion before execution
This profile is AI-generated and may contain inaccuracies.