Decart's Oasis 3 simulates hours of driving — but can it keep the world coherent?
decart released oasis 3, a photorealistic driving world model available via api for $0.02 per second. the model targets autonomous vehicle training but struggles with long-term consistency.
ai startup decart launched oasis 3, a real-time world model capable of generating photorealistic driving environments to train autonomous systems and robotics. built on the company's dos optimization stack, the model is priced at $0.02 per second and aims to foster a developer ecosystem similar to openai's strategy. the release follows a $300 million funding round that valued the company at nearly $4 billion, backing from investors including toyota and adobe. while the model demonstrates high photorealism and cost efficiency, early reviews indicate it degrades significantly during extended interactions, losing environmental consistency.