Alpamayo 1.5

Model Overview

Description:

Alpamayo 1.5 is a significant update to NVIDIA’s open 10B-parameter chain-of-thought reasoning VLA model, designed to be an interactive and steerable reasoning engine for the AV community. Alpamayo 1.5 is built on the Cosmos-Reason2 VLM backbone, is RL post-trained, and introduces support for navigation guidance, flexible camera counts, and user question answering.

This model is ready for non-commercial use.

License:

The model weights are released under a non-commercial license.

The inference code is released under the Apache 2.0 license.

Deployment Geography:

Global

Use Case:

Researchers and autonomous-driving practitioners who are developing and evaluating VLA models for autonomous-driving scenarios, particularly for handling rare, long-tail events.

Release Date:

Hugging Face 03/19/2026 via https://huggingface.co/nvidia/Alpamayo-1.5-10B

Inference Code:

GitHub: https://github.com/NVlabs/alpamayo1.5

Model Architecture:

Architecture Type: Transformer

Network Architecture: A VLA model based on Cosmos-Reason2 and featuring a diffusion-based trajectory decoder.

This model was developed based on: Cosmos-Reason2 (VLM backbone) with a diffusion-based action decoder

Number of model parameters:

Backbone: 8.2B parameters
Action Expert: 2.3B parameters

Input(s):

Input Type(s): Image/Video, Text, Egomotion History

Input Format(s):

Image: Red, Green, Blue (RGB)
Text: String
Egomotion History: Floating-point values (x, y, z), R_rot

Input Parameters:

Image: Two-dimensional (2D), multi-camera, multi-timestep
Text: One-dimensional (1D)
Egomotion History: Three-dimensional (3D) translation and nine-dimensional (9D, 3x3) rotation, multi-timestep

Other Properties Related to Input: Multi-camera images (4 cameras by default: front-wide, front-tele, cross-left, cross-right) with 0.4 second history window at 10Hz (4 frames per camera), image resolution 1080x1920 pixels (processor will downsample them to 320x576 pixels). Text inputs include user commands and navigation guidance. Images and egomotion history (16 waypoints at 10Hz) also require associated timestamps. Note that the model is primarily trained and only tested under this setting.

Output(s)

Output Type(s): Text, Trajectory

Output Format(s):

Text: String (Chain-of-Causation reasoning traces or question answers)
Trajectory: Floating-point values (x, y, z), R_rot

Output Parameters:

Text: One-dimensional (1D)
Trajectory: Three-dimensional (3D) translation and nine-dimensional (9D, 3x3) rotation, multi-timestep

Other Properties Related to Output: Outputs 6.4-second future trajectory (64 waypoints at 10Hz) with position (x, y, z) and rotation matrix R_rot in ego vehicle coordinate frame. Internally, the trajectory is represented as a sequence of dynamic actions (acceleration and curvature) following a unicycle model in bird's-eye-view (BEV) space. Text reasoning traces and question answers are variable in length, describing driving decisions and causal factors.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Runtime Engine(s):

PyTorch (minimum version: 2.8)
Hugging Face Transformers (minimum version: 4.57.1)
DeepSpeed (minimum version: 0.17.4)

Supported Hardware Microarchitecture Compatibility:

NVIDIA GPUs with sufficient memory to load a 10B parameter model (minimum 1 GPU with at least 24GB of VRAM)

Preferred/Supported Operating System(s):

Linux (we have not tested on other operating systems)

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

Alpamayo 1.5 10B trained

Can be integrated into autonomous driving software in the cloud for advanced end-to-end perception, reasoning, and motion planning.

Training, Testing, and Evaluation Datasets:

Training Dataset:

Alpamayo 1.5's training data comprises a mix of Chain of Causation (CoC) reasoning traces, Cosmos-Reason Physical AI datasets, NVIDIA's internal proprietary autonomous driving data, and public driving data.

Data Modality:

Image (multi-camera)
Text (reasoning traces, navigation guidance)
Other: Trajectory data (egomotion, future waypoints)

Image Training Data Size: More than 1 Billion Images (from 80,000 hours of multi-camera driving data)

Text Training Data Size: Less than a Billion Tokens (3M CoC reasoning traces, Cosmos-Reason training data, and public datasets)

Video Training Data Size: 10,000 to 1 Million Hours (80,000 hours)

Non-Audio, Image, Text Training Data Size: Trajectory data: 80,000 hours at 10Hz sampling rate

Data Collection Method by dataset: Hybrid: Automatic/Sensors (camera and vehicle sensors), Synthetic (VLM-generated reasoning)

Labeling Method by dataset: Hybrid: Human (structured CoC annotations), Automated (VLM-based auto-labeling and heuristic rules), Automatic/Sensors (trajectory and egomotion)

Properties: The dataset comprises 80,000 hours of multi-camera driving videos with corresponding egomotion and trajectory annotations. It includes 3,000,000 Chain-of-Causation (CoC) reasoning traces that provide decision-grounded, causally linked explanations of driving behaviors. Content includes machine-generated data from vehicle sensors (cameras, IMUs, and GPS) and synthetic reasoning traces. CoC annotations are in English and use a structured format that links driving decisions to causal factors. Sensors include RGB cameras (2-6 per vehicle), inertial measurement units, and GPS.

The training dataset also contains data from the following public datasets:

CODA-LM
Drive-Action
DriveGPT4
DriveLM
LingoQA
MapLM
MM-AU
NAVSIM-ReCogDrive
NAVSIM-Traj
nuInstruct
nuScenesQA
Omnidrive
Roadwork
Senna
SUTD
Talk2Car
W3DA

Testing Dataset:

Link: Proprietary autonomous driving test datasets, closed-loop simulation, on-vehicle road tests.

Data Collection Method by dataset: Hybrid: Automatic/Sensors (real-world driving data), Synthetic (simulation scenarios)

Labeling Method by dataset: Hybrid: Automatic/Sensors, Human (ground truth verification)

Properties: This dataset covers multi-camera driving scenarios with a particular focus on rare, long-tail events. It includes challenging cases such as complex intersections, cut-ins, pedestrian interactions, and adverse weather conditions. Data are collected from RGB cameras and vehicle sensors.

Evaluation Dataset:

Link: Same as Testing Dataset.

Data Collection Method by dataset: Hybrid: Automatic/Sensors (real-world driving data), Synthetic (simulation scenarios)

Labeling Method by dataset: Hybrid: Automatic/Sensors, Human (ground truth verification)

Properties: Evaluation focuses on rare, long-tail scenarios, including complex intersections, pedestrian crossings, vehicle cut-ins, and challenging weather and lighting conditions. Multi-camera sensor data are collected from RGB cameras.

Quantitative Evaluation Benchmarks:

Reasoning Evaluation using LingoQA: Lingo-Judge Score of 74.2.
Closed-Loop Evaluation using AlpaSim on 910 scenarios from the PhysicalAI-AV-NuRec Dataset: AlpaSim Score of 0.81 ± 0.01.
Open-Loop Evaluation on 937 challenging samples from the PhysicalAI-AV Dataset: minADE_6 at 6.4s of 1.11m.

Inference:

Acceleration Engine: PyTorch, Hugging Face Transformers

Test Hardware:

Minimum: 1 GPU with 24GB+ VRAM (e.g., NVIDIA RTX 3090, RTX 3090 Ti, RTX 4090, A5000, or equivalent)
Tested on: NVIDIA H100

For scripts related to model inference, please check out our code repository.

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.