Lumo-1 Context

Market context

Generalist VLA model designed to connect embodied reasoning with robot action generation.

Target environment: Robot-learning labs, Astribot S1 workflows and cross-embodiment research environments.
Workflow context: Vision-language-action reasoning, planning, trajectory prediction and action training on robot data.
Customer context: Robot-learning researchers and teams studying embodied reasoning for manipulation.
Deployment model: Research publication and model-development pathway.
Commercial maturity: Research-stage model described through a public paper rather than a standalone commercial product.
Adoption constraints: Adoption depends on model access, hardware compatibility, data quality, latency and real-task reliability.
Market position: Robot foundation-model layer for reasoning-to-action control.
Adjacent products: FluxVLA Engine, Gemini Robotics, LeRobot

Developers

Vision-language-action reasoning, planning, trajectory prediction and action training on robot data.

Robot-learning labs, Astribot S1 workflows and cross-embodiment research environments.

System Type: Vision-language-action robotics foundation model
Training Pipeline: Continued VLM pretraining; cross-embodiment co-training; action training; reinforcement learning
Control Scope: Reasoning-guided waypoint generation for robot manipulation
Sensing Modality: Vision and natural-language instruction
Robot Platforms Tested: Astribot S1 bimanual mobile manipulator
Evaluation Mode: Real-world pick-and-place, fine perception and long-horizon manipulation
Tasks Demonstrated: Generalizable pick-and-place and dexterous manipulation
Release Model: Research technical report