NVIDIA Launches Alpamayo 2 Super Open-Source Autonomous Driving Model
NVIDIA has officially released and commercialized Alpamayo 2 Super, a cutting-edge open-source inference model designed specifically for L4 autonomous vehicles and Robotaxis. The 34B parameter model combines the NVIDIA Cosmos 3 Super Reasoner backbone with a 2B parameter diffusion action expert, aiming to tackle complex real-world driving challenges.
Confirmed
- Model Parameters & Architecture: This Vision-Language-Action (VLA) model has a total of 34B parameters, integrating the Cosmos 3 Super Reasoner model with a 2B parameter diffusion action expert, and has been trained via reinforcement learning.
- Perception & Decision-making: It supports 7-camera 360-degree surround perception and possesses causal reasoning capabilities to formulate high-level driving decisions.
- Commercial & Open Source Status: The model is now officially available for commercial use in the autonomous vehicle and robotaxi sectors.
Why it matters
- The release and commercialization of this model provide the autonomous driving industry with a powerful, open-source, end-to-end VLA inference foundation, helping to accelerate the R&D and deployment of Robotaxis and related autonomous driving technologies.
2026-08-04 ~ 2026-08-04 · 7 related posts
Primary sources
- [source] NVIDIA Unveils Alpamayo 2 Super: 34B-Parameter AV Model with 7-Camera Surround Perception — NVIDIA Developer · 2026-08-04
- Alpamayo 2 Super Released: 34B Open Reasoning VLA for Autonomous Driving — tomaszbednarz · 2026-08-04
- NVIDIA Launches Alpamayo 2 Super: Open Reasoning Model for Autonomous Vehicles — MaziyarPanahi · 2026-08-04
- [source] Nvidia Releases Alpamayo 2 Super: A 34B Open VLA Model for Autonomous Driving — Scobleizer · 2026-08-04
3 near-duplicate retellings: nordicinst · nvidia · nvidia