Oru'el claims Kanook beats NVIDIA Flora and OpenAI on RoboRewardBench by 30-50%
AashaySachdeva · x · 2026-11-18
Startup Oru'el announced its robot reward-verification model Kanook as a new Tier-1 SOTA on RoboRewardBench.
- Kanook automates the annotation needed to judge whether a robot actually completed a task across every episode
- Per the shared chart, on 2,831 real robot episodes Kanook scores 27.98 MAE×100 (lower is better), 30% lower error than NVIDIA Flora-Rewarder (40) and 50% lower than OpenAI GPT-6 Astra (55.4)
- Other compared models include Gemini 3.1 Pro (99.6) and Claude Sonnet 5 (96.6), with scores cited from the Flora paper
Caveat: Oru'el is a very low-profile company and the chart lists unconfirmed model names like GPT-6 Astra and Kimi K3, so treat the claim with skepticism.
More from Embodied
- Voxel Robot Hails a Taxi, and Roboticists Are Excited — chris_j_paxton · 2026-11-18
- HT-Policy: multimodality isn't why diffusion policies win at imitation learning — chris_j_paxton · 2026-11-18
- Team builds on-device robot dog in 3 hours with phone-side Gemma, fully offline — DynamicWebPaige · 2026-11-18
- UMA Robotics bets on Europe first: survey shows EU edges US on talent, customers, cost — RemiCadene · 2026-11-18
- Tesla letter: FSD hits 1.74M paid users, Robotaxi in 7 cities ahead of Dec 3 vote — yunta_tsai · 2026-11-18
- Open-Source Robot Lamp Preps Skill Store: Install Physical Skills Like Apps — dee_hw · 2026-11-17