Generate Trajectories, Reasoning Traces, and Auto-Labels with NVIDIA Alpamayo 2 Super

Summary

Autonomous vehicle development typically relies on disjoint models for trajectory generation, intent prediction, scene understanding, and data labeling. NVIDIA Alpamayo 2 Super addresses this fragmentation with an open 34-billion-parameter vision-language-action model that couples a 32-billion-parameter Cosmos 3 Super Reasoner with a 2-billion-parameter diffusion-based Action Expert. Processing surround-view video across up to seven cameras, the model simultaneously produces future vehicle trajectories, Chain-of-Causation reasoning traces, high-level meta-actions, and grounded scene question answers. In benchmark evaluations, the model achieved a 6.4-second minADE_6 of 0.911 meters on the Physical AI AV Dataset and ranked first on LingoQA with a score of 79.2. Released under the OpenMDW-1.1 permissive license, the architecture serves as an offline policy teacher, evaluation critic, or auto-labeling engine that can be distilled for NVIDIA DRIVE AGX Thor deployment.

Takeaways

1. 1

   Alpamayo 2 Super combines a 32B Cosmos 3 Super Reasoner and a 2B diffusion Action Expert to jointly output future driving trajectories and Chain-of-Causation reasoning traces.

2. 2

   The model achieves a 6.4-second minADE_6 of 0.911 m on the Physical AI AV Dataset, a 0.433 AV reasoning benchmark score, and ranks first on LingoQA with a score of 79.2.

3. 3

   Released under the permissive OpenMDW-1.1 license, the model can generate auto-labels with or without ground-truth future trajectories and be distilled for NVIDIA DRIVE AGX Thor hardware.