The company says Dream-1 can operate a new bimanual robot with parallel grippers without fine-tuning, demonstrations collected from that robot, or machine-learning expertise from the user. Its stated aim is zero-shot generalization across unseen robots, environments, and tasks.
Dream Labs frames the release around cheaper hardware, stating that a capable bimanual arm setup can cost as little as $2,000. It also says API access is available worldwide and presents Dream-1 as a way to start using a newly connected robot immediately.
The source does not provide an API price, a full compatibility list, or independent measurements of task success, reliability, safety, or performance on robots beyond the company’s examples. It does not establish that the model will generalize to all bimanual systems or operating conditions.