๐ฐTechCrunch AIโขStalecollected in 23m
Decart launches Oasis 3 for real-time driving simulation
๐กAccess a real-time world model API to simulate complex driving scenarios for autonomous vehicle development.
โก 30-Second TL;DR
What Changed
Oasis 3 generates photorealistic, real-time driving environments.
Why It Matters
This release lowers the barrier for developers to create high-fidelity synthetic data for autonomous driving. It could significantly reduce the reliance on expensive physical road testing for edge-case scenarios.
What To Do Next
Sign up for the Decart API to integrate Oasis 3 into your autonomous vehicle simulation pipeline for stress-testing edge cases.
Who should care:Developers & AI Engineers
Key Points
- โขOasis 3 generates photorealistic, real-time driving environments.
- โขThe model is designed specifically for autonomous vehicle testing.
- โขDevelopers can access the simulation capabilities via a new API.
- โขThe release aims to accelerate the training and validation of self-driving AI.
๐ง Deep Insight
Web-grounded analysis with 18 cited sources.
๐ Enhanced Key Takeaways
- โขOasis 3's launch is backed by a recent $300 million Series C funding round in May 2026, valuing Decart at approximately $4 billion, with participation from investors like Nvidia, Adobe Ventures, and Toyota Ventures.
- โขThe release of Oasis 3 coincides with the introduction of the Decart Optimization Stack (DOS) 2.0, which is designed to enhance high-performance inference capabilities for low-latency AI workloads.
- โขDecart strategically positions Oasis 3 as an infrastructure product, offering API access for developers to build on top of it, rather than creating end-to-end autonomous vehicle simulation platforms themselves.
- โขOasis 3 evolved from earlier Oasis versions, first made public in October 2024, which were interactive open-world models initially demonstrating real-time generative AI video for gaming (Minecraft-like environments).
- โขDecart also offers the Lucy model for immersive experiences, indicating a broader portfolio beyond just driving simulation, with Oasis specifically targeting physical AI applications like robotics and autonomous vehicles.
๐ Competitor Analysisโธ Show
| Company/Product | Key Features | Target Industry | Pricing/Benchmarks |
|---|---|---|---|
| Decart Oasis 3 | Photorealistic, real-time driving environments; API access; leverages Decart Optimization Stack (DOS) 2.0; evolved from general world models. | Autonomous Vehicle Testing, Robotics, Physical AI | $0.02 per second for API access; enterprise pricing varies. |
| Siemens Simcenter Prescan | Physics-based simulation; digital twin technology; scenario-driven testing; high-fidelity sensor modeling (radar, lidar, camera, V2X); MiL, SiL, HiL, DiL, cloud deployments; MATLAB/Simulink integration. | Autonomous Vehicles, ADAS | Not publicly disclosed. |
| Ansys Autonomous Vehicle Simulation | High-fidelity simulation; multiphysics capabilities; environment modeling; sensor simulation; virtual prototyping; testing of perception algorithms, decision-making logic, and control systems. | Autonomous Vehicles, ADAS | Not publicly disclosed. |
| NVIDIA DRIVE Sim | GPU-accelerated simulation; AI-enabled; realistic, data-rich test environments for developing, training, and validating autonomous systems. | Autonomous Vehicles, ADAS | Not publicly disclosed. |
| MathWorks (MATLAB/Simulink) | Model-based design tools; flexible architecture supporting hardware-in-the-loop (HiL) and software-in-the-loop (SiL) approaches; rapid prototyping and validation. | Autonomous Systems Development, General Simulation | Not publicly disclosed. |
๐ ๏ธ Technical Deep Dive
- Architecture: Oasis 3 is based on a Diffusion Transformer model.
- Components: It comprises a spatial autoencoder (based on Vision Transformer, ViT) and a latent diffusion backbone (based on Diffusion Transformer, DiT).
- Generation Method: Generates frames autoregressively, conditioning each frame on user input for real-time interaction.
- Temporal Stability: Employs "dynamic noising" to adjust inference-time noise, reducing error accumulation and maintaining visual consistency over long time horizons.
- Performance Optimization: Optimized for Etched's Sohu Transformer ASIC, aiming for 4K resolution and 100 billion parameter support, though currently runs on Nvidia H100 GPUs.
- API Access: Provides SDKs for JavaScript, Python, Swift, and Android, with Realtime, Queue, and Process APIs for various use cases including live stream transformation, video processing, and image editing.
- Previous Performance (Oasis 1.0/Minecraft demo): Achieved 20 frames per second (fps) at 460p with near-zero latency (0.04 seconds per frame) on Nvidia H100 GPUs.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Decart's Oasis 3 will significantly accelerate the development and validation cycles for autonomous vehicle (AV) systems.
By providing photorealistic, real-time simulation environments via an accessible API, developers can rapidly test a wider range of edge cases and dangerous scenarios virtually, reducing reliance on costly and time-consuming physical testing.
Decart is poised to become a foundational infrastructure provider for the broader "physical AI" and robotics industries.
By offering Oasis 3 as an API-first world model and having a history of general-purpose world models (Oasis 1.0, Lucy), Decart is positioning itself as a core technology layer for various real-time interactive AI applications beyond just AVs.
The adoption of world models like Oasis 3 will drive increased demand for specialized AI hardware accelerators.
The high computational demands of real-time, photorealistic world models necessitate powerful inference capabilities, as evidenced by Decart's optimization for Etched's Sohu ASIC and current reliance on Nvidia H100 GPUs.
โณ Timeline
2022
Decart co-founders Dean Leitersdorf and Moshe Shalev meet, conceiving the idea for Oasis.
2023
Decart is founded.
2024-10-31
Decart publicly releases Oasis (Minecraft-like demo), an interactive open-world AI model.
2025-08
Decart raises a $100 million Series B funding round, achieving a $3.1 billion valuation.
2026-05
Decart completes a $300 million Series C funding round, valuing the company at $4 billion, and introduces the Decart Optimization Stack (DOS) 2.0.
2026-06-10
Decart launches Oasis 3, a world model for photorealistic real-time driving simulation, accessible via API.
๐ Sources (18)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: TechCrunch AI โ