Robocurve evaluations
Real-world evaluations of physical AI that score robot and vision-language-action models on its RoboDojo tasks and language models on refusing harmful robot instructions, publishing mean score with estimated cost and time per attempt; results are public and labs of all sizes can ask the company to evaluate their model and robot; engagements are arranged by contacting the company and no prices are published.
Find alternatives to Robocurve evaluations
What it does
- Evaluation and observability
Features 4
- Real-world robot scoresPublishes real-world scores on RoboDojo-RC Tier 1 for language models and robot models.
- Harmful-instruction refusalsPublishes refusal rates on RoboHarm v1.
- Cost per attemptReports estimated cost per task attempt for each model configuration.
- Evaluation for labsOffers to evaluate a lab's model and robot.
Sources
- Pricing
-
No pricing page is published; the home page offers Talk to us and a contact form listing Evaluation as an area of interest. The dollar figures on the page are estimated cost per benchmark attempt, not prices.
- Description
-
Official documentation · 20 Oct 2026
Home page: "Evaluate your model and robot" and "We work with labs of all sizes".
- What it does
-
Official documentation · 20 Oct 2026
Home page: "Evaluate your model and robot" and "We work with labs of all sizes".
- Sold within
-
Results are published openly; evaluation engagements are arranged through a contact form and no prices are published.
- Name
-
Official documentation · 20 Oct 2026
Name as published on the product page.
- URL
-
Official documentation · 20 Oct 2026
Product page, read 2026-10-20.
- Deployment
-
Official documentation · 20 Oct 2026
The pages read state no deployment model beyond the vendor-run service; saas is recorded as the vendor-operated reading.
Open Robocurve in the directory
Something wrong here? Send a correction — quote this product id: robocurve-evaluations.