7-month-old SynapX details embodied world model SYNWorld V0.5, 60x faster inference beats π0.5
AGI Hunt · wechat · 2026-10-21
SynapX, a 7-month-old robotics startup founded by Horizon Robotics' 6th employee, published a full technical blog on its embodied world model SYNWorld V0.5, outperforming π0.5 and NVIDIA Cosmos 3 Nano in benchmarks.
Key details: a Mixture-of-Transformers architecture trained on 300k+ hours of data (only 10% real-robot) unified via a 59-dim action space spanning human video, Ego, UMI and robot data; EMG-wristband data capture; inference optimizations (FP8, two-step DMD2 distillation) cutting full denoising from 7.9s to 131ms — 60x faster, enabling real-robot deployment. It beats π0.5 in all 16 real-task comparisons (advantage largest with few demos) and scores 87.78% on LIBERO-Plus vs Cosmos 3 Nano's 85.66%. It ships alongside the 23-DoF OctoH-Hand dexterous hand with 1900+ tactile sensors, forming a brain-hand-data loop.
More from Embodied
- MIT CSAIL's X-Hinges 3D-Prints Everyday Objects With Built-In Sensing Across Three Motion Axes — MIT News AI · 2026-10-23
- Formic raises $50M Series B to scale physical AI across 100+ US factories — garrytan · 2026-10-23
- fal starts fine-tuning a physical robot to draw, results next week — OdinLovis · 2026-10-23
- PEARL demo: an LLM agent that checks experiment results, revises plans, and re-tests — KevinKaichuang · 2026-10-23
- CU Boulder's ATLAS Institute Opens Tenure-Track Robotics Professorship — ryosuzk · 2026-10-23
- MiTaS multi-resolution tactile imitation learning hits 80% success vs 35.5% vision-only baseline — Jan_R_Peters · 2026-10-23