For robotaxis and other autonomous vehicles (AVs), the most challenging scenarios are often rare, complex ‘long-tail’ events that are difficult to anticipate, reproduce and train for.
Nvidia’s Alpamayo 2 Super targets the complex long-tail scenarios that remain challenging for autonomous vehicles. The model enables AVs to reason about situations, assess cause and effect, select actions and generate safe driving paths in real time. Designed to support inspectable and verifiable decision-making, Alpamayo 2 Super is now available for commercial use as part of Nvidia’s open Alpamayo family of autonomous driving foundation models.
Built on Nvidia Cosmos 3 Super Reasoner and post‑trained with reinforcement learning, the model advances the AV ecosystem on two fronts: open commercial licensing and leading multitask capabilities for autonomous driving.
Open licensing for production AVs
Alpamayo 2 Super is available on Hugging Face under the Linux Foundation’s permissive OpenMDW license, allowing AV developers, auto makers and suppliers to fine-tune, create derivative models and commercially deploy the technology using their own data and infrastructure. The approach provides a path from model adaptation to deployment while allowing developers to retain control of proprietary data, specialized models and safety-related workflows.
Nvidia’s Alpamayo model family supports a cloud-to-vehicle workflow for autonomous driving development. Alpamayo 2 Super provides high-level multimodal reasoning for generating training data and model outputs, while Alpamayo 1.5 and Alpamayo 1 offer more cost-efficient development options. Resulting models can then be distilled and optimized for real-time inference in production vehicles, supporting scalable deployment across autonomous fleets.
Benchmark-leading reasoning at frontier scale
Nvidia says Alpamayo 2 Super ranks first on LingoQA, an autonomous driving reasoning benchmark, among nearly 40 evaluated models. In Nvidia’s testing, it outperformed Qwen2.5-VL 72B, Gemini 2.5 Pro and GPT-4o on the Lingo-Judge metric, while also ranking first across the autonomous driving benchmarks assessed by the company.
Nvidia’s Alpamayo 2 Super has three times the scale of its 10-billion-parameter Alpamayo 1.5 and Alpamayo 1 models, supporting improved reasoning from sparse examples. The model processes full-surround camera data to provide 360° context, helping autonomous vehicles interpret complex scenarios including lane changes, merges, unprotected turns and challenging intersections.
Nvidia’s Alpamayo 2 Super is designed as a multitask foundation model for robotaxis and autonomous driving, producing five outputs for each driving situation: a planned trajectory, a chain-of-causation reasoning trace, a meta-action representing driving intent, automated reasoning labels for training and validation, and visually grounded answers linked to specific regions of camera images.
Together, these outputs offer insight into the model’s decision-making process. Developers can tie what the model observed to the action it selected, making decisions easier to understand, critique and validate.
CoC traces integrate with Nvidia Halos safety‑validation workflows and support AI safety aligned with ISO/PAS 8800 requirements, providing a stronger foundation for AV safety engineering.
Alpamayo 2 Super can also be deployed as an auto-labeler to generate CoC labels and perform visual question answering with 2D grounding on proprietary fleet data. By linking its reasoning to specific regions in camera images, the model can transform raw driving clips into richer training data, compressing annotation cycles from months to days.
Beyond planning and auto-labeling, Alpamayo 2 Super supports scene understanding, model critiquing and knowledge distillation. These multitask capabilities enable developers to use a single foundation model across more of the development stack, simplifying tooling and accelerating iteration.
Alpamayo has already surpassed 500,000 downloads on Hugging Face.
In related news, Toyota builds AI-defined mobility platform with Nvidia