Skip to content
Trending storyDeveloping

ModelScope releases uni-robot-agent harness, reports open-weight Qwen matching GPT-6 Astra

1 article1 sourcesince Nov 3Last article 1h ago ·

Overview

AISummary of 1 article

ModelScope releases uni-robot-agent, a unified harness for robot-use agents, and says it runs the open-weight Qwen3.8-Flash-Next model without privileged simulator inputs.

On the RoboDojo benchmark the company reports a score of 31.39 with 25.00% success, against 28.97 and 22.48% for the published GPT-6 Astra–RoboProbe configuration.

Across 19 physical-robot tasks, ModelScope says the zero-shot Qwen agent scored 49.50/100, ahead of fine-tuned π0.5 at 47.75 and zero-shot Dream-1 at 35.50. These figures come from the company's own post, and no independent evaluation is cited.

ModelScope says shared motion, perception, skill, and episode interfaces make trajectories across environments and robot embodiments easier to combine for future training.

Written by AI from the articles below · updated Nov 3, 10:13 AM ET

Check the sources:

Latest developmentNov 3, 10:11 AM ET
ModelScope says open-weight Qwen matches GPT-6 Astra on robot-use benchmark

Article timeline

The articles in this story. Times are ET.

Nov 3
  1. ModelScopeOfficial
    ModelScope says open-weight Qwen matches GPT-6 Astra on robot-use benchmark

    AIModelScope says its uni-robot-agent harness, running Qwen3.8-Flash-Next, scored 31.39 with 25.00% success on RoboDojo without privileged simulator inputs. That compares with 28.97 and 22.48% for the published GPT-6 Astra–RoboProbe configuration. Across 19 physical-robot tasks, the zero-shot Qwen agent scored 49.50/100, ahead of fine-tuned π0.5 at 47.75 and zero-shot Dream-1 at 35.50, according to the post.

    Image from @ModelScope2022's post

Heat Trend

Not enough continuous observations to show a trend yet.