DYNA-1
About
About DYNA-1
DYNA-1, also called Dynamism v1, is Dyna Robotics’ first robot foundation model for continuous dexterous manipulation. Its training loop uses a learned reward model to score robot behaviour, guide online exploration and select useful experience for further training. The model was developed around contact-rich work with deformable objects, where recovery from folds, occlusion and changing object geometry matters during long autonomous runs.
Dyna reports that DYNA-1 folded more than 850 napkins over 24 hours with 99.4% task success and no human intervention, operating at about 60% of human speed. The company ran the system in a restaurant environment and also demonstrated manipulation of shirts and cup-filling tasks. DYNA-1 is the learned control model used with robot hardware; it is separate from Dyna’s later DYNA-2 model and DYNA-SAUR physical embodiment.
Timeline
- Din Tai Fung begins network rollout of Dyna robots
2026-08-27 - Deployment - Dyna-2 folds 95 napkins per hour, with 93% clearing the restaurant's quality threshold and 1,590 table-ready napkins produced per 18-hour day. The customer target is 1,500 table-ready napkins per robot per 18-hour shift; Dyna-1 previously produced about 480 at 35 folds per hour and 75% quality. Rollouts across restaurants, hotels, logistics sites and data centres are expected to take Dyna's total deployed fleet into the hundreds by the first half of 2027.
Source - Dyna Robotics unveils DYNA-1 foundation model
2025-04-29 - Product Launch - The company described DYNA-1 as designed for sustained round-the-clock dexterous operation. Dyna said the model supported commercial tasks across folding and food-preparation workflows. The product gives Dyna Robotics a concrete offering for commercial automation and autonomous-system workflows.
Source