Decart Launches Oasis 3 AV Simulator at $0.02/Sec
Key insights
- Decart raised $300 million at a nearly $4 billion valuation with Toyota, Adobe, eBay, and Nvidia as named strategic investors.
- Each frame generates roughly 8,000 tokens, filling context windows quickly and degrading scenes into generic urban environments during extended sessions.
- Oasis 3 lacks physics simulation: vehicles drive through other cars and directional controls are frequently unresponsive.
Why this matters
Decart's $0.02-per-second price point could make high-volume, rare-event simulation accessible to AV teams that cannot afford purpose-built simulation infrastructure, directly lowering the cost of edge-case training data generation. Toyota and Nvidia joining as strategic investors in a nearly $4 billion company signals that major automotive and chip players are placing real bets on generative world models as production AV development infrastructure. The absent physics simulation and context windows filling at roughly 8,000 tokens per frame are the central issues determining whether Oasis 3 can enter any regulatory or pre-deployment AV testing pipeline.
Summary
Decart unveiled Oasis 3, a real-time world model for photorealistic driving scenarios, available at $0.02 per second.
Backed by $300 million at a nearly $4 billion valuation, with Toyota, Adobe, eBay, and Nvidia as investors, CEO Dean Leitersdorf says Oasis 3 is more than an order of magnitude cheaper than anyone else in the industry.
Essentially: (Decart) is positioning Oasis 3 as AV and robotics simulation infrastructure.
- Each frame generates roughly 8,000 tokens, filling context windows fast and causing scenes to degrade into generic urban environments.
- Physics is absent: vehicles drive through other cars and directional controls are frequently unresponsive.
With more than 100,000 developers on its Lucy model, Decart has distribution, but physics and memory gaps block production AV use.
Potential risks and opportunities
Risks
- AV teams that adopt Oasis 3 for safety-critical edge-case simulation risk producing misleading training datasets if physics-free vehicle interactions pass internal review without rigorous downstream filtering
- If a competitor ships a world model with physics simulation within the next 12 months, Decart's nearly $4 billion valuation faces pressure given Oasis 3's documented physics and memory gaps
- Toyota and Nvidia, named as strategic investors, carry reputational exposure if Oasis 3 is adopted in any pre-deployment AV safety pipeline before physics and context-window limitations are resolved
Opportunities
- Physics-based AV simulation providers gain a clear differentiation argument as Oasis 3's physics-free outputs make the case for hybrid workflows that layer generative world models over traditional physics simulators
- Decart's more than 100,000 developer ecosystem on its Lucy video model gives it a ready channel to pull robotics and physical AI teams toward Oasis 3 ahead of competitors reaching general availability
- Nvidia, named as both a hardware partner and strategic investor, is positioned to deepen AV infrastructure lock-in if Oasis 3 adoption grows across teams already running on Nvidia hardware
What we don't know yet
- Whether Decart's extended memory research has a timeline for resolving the context window exhaustion that causes previously visited locations to be replaced rather than remembered
- How Toyota plans to integrate Oasis 3 into its AV development workflow, and whether a commercial supply agreement exists beyond the strategic investment
- What specific competitors and their prices Decart benchmarks against when CEO Dean Leitersdorf claims more than an order of magnitude cheaper than anyone else in the industry
Originally reported by techcrunch.com
Read the original article →Original headline: Decart Launches Oasis 3 World Model for AV Simulation — Hours of Photorealistic Driving at $0.02/Second With Physics Caveats