← Leaderboard

SenseNova-RoboRSI

Agent system
SenseTime · Oct 15, 2026

Agentic system from the SenseNova joint research team that uses GPT-6 Astra directly as the policy, labelled “Agentic LM” on RoboLab-120 (which credits Nanyang Technological University). Code release is announced but pending.

Type
Agent system
Parameters
—
Backbone
GPT-6 Astra
Weights
Unknown
License
–
Commercial use
Unclear
Released
Oct 15, 2026
Robots
–
SO-101
–
LeRobot
–

Benchmark results

2 of 13 benchmarks
RoboLab-120Simulation
#2 of 2257.1%
#5 of 4049.2%
Not reported on 11 benchmarks

Adoption

updated Nov 17, 2026
GitHub stars
825
GitHub forks
90

Speed per task

SettingPer taskPer planner callCallsCostSource
RoboDojo sim, GPT-6-Astra + ICL, 210 eps, 1× budget · GPT-6-Astra API54.9 min–––Oct 2026 ↗
LIBERO-PRO Task/Swap, 800 eps, 1× budget · GPT-6-Astra API (max reasoning)19 min–––Oct 2026 ↗
LIBERO-Long++, 120 eps, 1× budget · GPT-6-Astra API (max reasoning)53.1 min–––Oct 2026 ↗
LIBERO-Ultra-Long, 90 eps, 1× budget · GPT-6-Astra API (max reasoning)1.8 h–––Oct 2026 ↗
Wrong or missing number? Report an error