NVIDIA today announced a significant expansion of the Alpamayo open platform, designed to support the development of reasoning-based autonomous vehicles (AVs). The platform provides researchers and developers with a flexible, high-performance, and scalable suite of models, datasets, simulation, and training tools for building and evaluating modern AV stacks in realistic closed-loop settings. The Alpamayo open platform, introduced at CES 2026 and expanded at GTC 2026, has seen rapid adoption across industry and academia, with its reasoning models collectively surpassing 400,000 downloads to date. The platform was also recognized with a Computex 2026 Best Choice Award in the Vehicle Technology & Smart Cockpit category for its contributions to open, reasoning-based AV development.
The new additions to the Alpamayo platform include the Alpamayo 2 Super model, which is built on the Cosmos 3 Super Reasoner 32-billion parameter VLM backbone, offering 3x the parameter scale of previous models. This model introduces support for surround-view camera inputs, reasoning autolabeling, 2D grounding, and meta-action outputs. It also provides state-of-the-art performance in reasoning quality, trajectory accuracy, alignment, and more. Additionally, the platform now includes AlpaGym, an open-source, high-throughput closed-loop reinforcement learning (RL) framework that enables efficient, scalable, closed-loop RL for driving performance. AlpaGym runs models through continuous decision and observation cycles in AlpaSim, exposing compounding errors and edge-case failures that static datasets miss.
NVIDIA also announced the release of two Hugging Face challenges to benchmark closed-loop driving behavior and reasoning capability. The AlpaSim End-to-End Closed-Loop Challenge evaluates submitted driving policies in closed-loop, measuring how long models can drive without at-fault incidents in reconstructed real-world scenarios. The Physical AI AV Reasoning Challenge invites the research community to build models that can reason about long-tail scenarios in natural language. These challenges aim to bolster new innovations in the field and accelerate the development of level 4 AV systems.
Source: huggingface