Waddle-1 robot-use agent scores 81 on unseen bimanual tasks, beating GPT-6.2 Astra at 66
ycombinator · x · 2026-11-21
Waddle Labs announced Waddle-1, its first robot-use agent that controls bimanual robots zero-shot on unseen tasks. On the real-world RoboDojo-RC Tier 1 benchmark it scored 81 vs 66 for GPT-6.2 Astra and 59 for Dream-1, also outperforming Gemini 4, Opus 5.5, and NVIDIA GR00T N1.7.
Architecturally, Waddle-1 calls cloud LLMs for high-level reasoning while lower-level reactive policies run locally, aiming to combine reasoning models' capability with reactive policies' efficiency. The team expects such systems to reach human-speed manipulation soon and is taking preview API signups.
More from Embodied
- Neuralink patient plays GTA VI with brain implant, Shivon Zilis celebrates — shivon · 2026-11-21
- FlowPilot wins Best Paper at CoRL 2026 workshop for preference-based sidewalk navigation — kaiwei_chang · 2026-11-21
- Dev builds Iron Man to Dr. Doom AR lens switch in one day with Claude — Scobleizer · 2026-11-21
- VR-teleop-kit ported to Ulixarm robot arm by community developer — DominiqueCAPaul · 2026-11-21
- Agentic Robotics Night in Zürich: founders panel on building robots that perceive, reason and act — DominiqueCAPaul · 2026-11-21
- RoboWild benchmark: best frontier model scores just 13.3% on real-world robot use across 341 tasks — taoyds · 2026-11-21