# RoboFYI > Compare robot foundation models (VLAs, world-action models and agent systems) across simulation, real-robot and low-cost-arm benchmarks. Every score links to its source. RoboFYI tracks 140 robot foundation models across 13 benchmarks (simulation, real robot and low-cost arms), plus 37 datasets and 35 robots. Data last updated 2026-11-17. How to read scores: values are success rates in % unless the benchmark's unit says otherwise (RoboArena is a rating). Provenance per score: "official" (benchmark leaderboard entry), "self" (reported by the model's authors), "external" (reported by a third party), "unverified" (seen second-hand). Ranks are competition ranks over all models with a score. Agent systems (an LLM planner driving policies) are typed "agent" or "llm". ## Data (JSON, no auth, CORS-friendly static files) - [leaderboard.json](https://robofyi.com/data/leaderboard.json): every benchmark with its full ranking (rank, model, score, provenance, source URL) - [results.json](https://robofyi.com/data/results.json): every individual result, including non-standard and self-reported runs, with notes - [models.json](https://robofyi.com/data/models.json): model metadata: lab, type, size, backbone, weights, license, robots, links - [benchmarks.json](https://robofyi.com/data/benchmarks.json): what each benchmark measures, unit, saturation, leaderboard URL - [datasets.json](https://robofyi.com/data/datasets.json): robot learning datasets: size, robots, license, which models train on them - [hardware.json](https://robofyi.com/data/hardware.json): robots: price, DoF, payload, reach, which models and benchmarks use them - [speed.json](https://robofyi.com/data/speed.json): inference latency per action chunk and wall-clock time per task - [llms-full.txt](https://robofyi.com/llms-full.txt): every leaderboard as a Markdown table ## Pages - [Leaderboard](https://robofyi.com): all models, sortable by benchmark; URL parameters: sort=, view=sim|real|lowcost, types=vla,wam,policy,agent,llm, official=true - [Benchmarks](https://robofyi.com/benchmarks): what each benchmark measures and who leads it - [Charts](https://robofyi.com/charts): size vs score, speed vs score, sim vs real, progress, open vs closed - [Datasets](https://robofyi.com/datasets) and [Hardware](https://robofyi.com/hardware): catalogs with detail pages - Model pages: https://robofyi.com/models/ (ids in models.json) ## Optional - [Sitemap](https://robofyi.com/sitemap.xml) - [Report an error or submit a result](https://robofyi.com/report)