Is SenseNova-RoboRSI-27B-Preview state of the art?Pending benchmarks.Announced Oct 19, 2026; no independent results yet.
SenseTime's 27B embodied model for robot planning and decision-making, trained from Qwen3.8-27B with experience from its open-source SenseNova-RoboRSI harness, which runs and improves robot agents in simulation and on real robots. Released alongside SenseTime's study of GPT-6 Astra on 12 simulation benchmarks and 8 real robots. Open weights announced, not yet public.
The details so far
- Lab
- SenseTime
- Announced
- Oct 19, 2026
- API
- Not yet
- Open weights
- Announced, not yet published
- Parameters
- 27B (Qwen3.8-27B base)
- Base model
- Qwen3.8-27B, trained with recursive self-improvement (RSI) on validated interaction experience
- Harness
- SenseNova-RoboRSI on GitHub (Apache 2.0): shared runtime for LIBERO, LIBERO-PRO, LIBERO-Long++, Bench2Dex, DexJoCo and BEHAVIOR-1K, with code still rolling out
- Execution
- Direct VLM control (System 2) or VLA-assisted (System 2 + System 1)
- Weights
- Announced for Hugging Face (sensenova/sensenova-roborsi-27b-preview; README said Oct 20), not public as of Oct 21; technical report to follow
Tags marked unsure are not stated on the lab's own pages.
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab claim: runs at about 1/50 the cost of GPT-6 Astra.
Lab claim: 26.13 average on RoboDojo (2.84 below GPT-6 Astra), 95.0% on LIBERO-Pro with VLA-assisted execution and 96.1% with direct control, and RoboCasa365 success raised from 26.0% to 46.5% over the Qwen3.8-27B base.
Sources
LABSenseTime announcement on Xx.com/SenseTime_AI/status/2112166037936312671 · Oct 19, 2026REPOSenseNova-RoboRSI on GitHubgithub.com/OpenSenseNova/SenseNova-RoboRSI · Oct 21, 2026LABHas GPT-6 Astra achieved embodied intelligence? (SenseNova)sensenova.ai/has-gpt-6-astra-achieved-embodied-intelligence · Oct 19, 2026Tracking ends Jan 19, 2027 unless the API or weights become available.