Robot Software News
6 stories about robot software.
Alibaba Qwen released Qwen-VLA โ unified VLA model controlling humanoid robots, cross-embodiment โถ Video
Alibaba Qwen released Qwen-VLA, a unified Vision-Language-Action model integrating manipulation, navigation, trajectory prediction, and cross-embodiment control (single-arm, dual-arm, humanoid) into one system. Adapts via embodiment-aware prompts without separate training heads. 76.9% OOD success on ALOHA dual-arm tasks. Major LLM company entering Physical AI race โ joining Google Gemini Robotics and Mistral Physics AI.
X Square Robot open-sources Wall-OSS-0.5 VLA model under "Pretrain Once, Act Anywhere"
X Square Robot open-sourced Wall-OSS-0.5, a Vision-Language-Action model for real-world robotic manipulation. The pretrained checkpoint shows zero-shot generalization on multiple tasks without post-training, and outperforms open-source models like pi0.5 in controlled comparisons. Demonstrates that action-aware training improves robot-relevant understanding.
Genesis GENE-26.5 โ single model dexterous manipulation, 54 DoF bimanual hand, 200k hours training โถ Video
Genesis GENE-26.5 robotics model: single robotic hand with 54 DoF on bimanual system grasping multiple objects at once. Trained on 200,000+ hours real human hand data. Zero fine-tuning across sequences. Scaling data 4x raised success from 16.6% to 65.6% on long-horizon tasks.
Zurich is Europe's robotics capital โ ecosystem deep-dive by Lukas Ziegler
Zurich's robotics ecosystem: ETH Zurich talent, ABB proximity, deep-tech funding. 20+ companies highlighted: ANYbotics (50M), Voliro (3M), Flexion Robotics (0M), Ascento, Wingtra, Auterion (30M), and many more.
Newton โ open-source physics simulator by NVIDIA, DeepMind, Disney (Linux Foundation)
Newton is an open-source physics simulator for robotics, co-developed by NVIDIA, Google DeepMind, and Disney Research, now under the Linux Foundation (Apache 2.0). Runs thousands of robots in parallel on one GPU. Up to 475x faster than MJX for manipulation.
Eka Robotics unveils Vision-Force-Action Model for robot manipulation โถ Video
Eka Robotics, founded by MIT's Pulkit Agrawal, announces their Vision-Force-Action (VFA) model โ treating force sensing as a first-class modality alongside vision for robotic manipulation. The approach aims to give robots a native understanding of physical forces.
Get Robot Software Updates
Subscribe to receive robot software news directly in your inbox.
Subscribe