[HUGGINGFACE]score: 0.84

Open-Weight Qwen3.8-Flash-Next Outperforms GPT-6 Astra on Robot Tasks

October 31, 2026

The uni-robot-agent harness enables training for open-weight robot-use agents. Qwen3.8-Flash-Next achieved a RoboDojo Score of 31.39, surpassing the GPT-6 Astra-RoboProbe result of 28.97 on 19 real-world tasks.

HOW THIS AFFECTS YOU

●

builderYou can now train open-weight models for robot manipulation using a unified harness.

●

researcherThis provides a competitive baseline for VLA/WAM research without proprietary models.

●

founderOpen-weight models are now competitive with frontier proprietary models in robotics.

read original ↗huggingface.co

DAILY DIGEST

you don't check 9 sources — we do. one email every morning, read in 2 min. free. unsubscribe anytime. privacy

← back to feed