Sold by: NVIDIA
Deployed on AWS
Vision language model that excels in understanding the physical world using structured reasoning on videos or images.
Overview
NVIDIA Cosmos Reason 2 is an open, customizable, 8B-parameter reasoning vision language model (VLM) for physical AI and robotics that enables robots and vision AI agents to reason like humans, using prior knowledge, physics understanding and common sense to understand and act in the real world. This model understands space, time, and fundamental physics, and can serve as a planning model to reason what steps an embodied agent might take next.
New features with Cosmos Reason 2:
- Enhanced physical AI reasoning with improved spatio-temporal understanding and timestamp precision.
- Supports object detection with 2D/3D point localization and bounding box coordinates with reasoning explanations and labels. Improved long-context understanding up to 256K input tokens.
Use cases:
- Video analytics AI agents: Extract valuable insights and perform root-cause analysis on massive volumes of video data. These agents can be used to analyze and understand recorded or live video streams across city and industrial operations. Jumpstart your development of video analytics AI agents by using the NVIDIA Blueprint for video search and summarization (VSS) with Cosmos Reason as the VLM.
- Data curation and annotation: Enable developers to automate high-quality curation and annotation of massive, diverse training datasets. Experience NVIDIA Cosmos Curator, powered by Cosmos Reason, a framework that enables developers to quickly filter, annotate, and deduplicate large amounts of sensor data necessary for physical AI development.
- Robot planning and reasoning: Act as the brain for deliberate, methodical decision-making in a robot vision language action (VLA) model. Now robots such as humanoids and autonomous vehicles (AV) can interpret environments and complex commands, break them down into tasks and execute them using common sense, even in unfamiliar environments. Explore the NVIDIA Isaac GR00T-Dreams blueprint, which generates vast amounts of synthetic trajectory data using NVIDIA Cosmos world foundation models. Explore the Cosmos Cookbook, a technical guide that delivers end-to-end workflows, implementation recipes, and detailed examples for building, fine-tuning, and deploying Cosmos Reason in production-ready environments.
The model is ready for commercial use.
Highlights
- Architecture Type: A Multi-modal LLM consists of a Vision Transformer (ViT) for vision encoder and a Dense Transformer model for LLM.
- Cosmos-Reason2-8B is post-trained based on Qwen3-VL-8B-Instruct and follows the same model architecture.
- Number of model parameters: Cosmos-Reason2-8B: 8,767,123,696
Details
Sold by
Categories
Delivery method
Latest version
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.g5.48xlarge Inference (Batch) Recommended | Model inference on the ml.g5.48xlarge instance type, batch mode | $8.00 |
ml.g6e.xlarge Inference (Real-Time) Recommended | Model inference on the ml.g6e.xlarge instance type, real-time mode | $1.00 |
ml.g5.48xlarge Inference (Real-Time) | Model inference on the ml.g5.48xlarge instance type, real-time mode | $8.00 |
ml.g6e.2xlarge Inference (Real-Time) | Model inference on the ml.g6e.2xlarge instance type, real-time mode | $1.00 |
ml.g6e.4xlarge Inference (Real-Time) | Model inference on the ml.g6e.4xlarge instance type, real-time mode | $1.00 |
ml.g6e.8xlarge Inference (Real-Time) | Model inference on the ml.g6e.8xlarge instance type, real-time mode | $1.00 |
ml.g6e.16xlarge Inference (Real-Time) | Model inference on the ml.g6e.16xlarge instance type, real-time mode | $1.00 |
ml.g6e.12xlarge Inference (Real-Time) | Model inference on the ml.g6e.12xlarge instance type, real-time mode | $4.00 |
ml.g6e.24xlarge Inference (Real-Time) | Model inference on the ml.g6e.24xlarge instance type, real-time mode | $4.00 |
ml.g6e.48xlarge Inference (Real-Time) | Model inference on the ml.g6e.48xlarge instance type, real-time mode | $8.00 |
Dimensions summary
You pay by the host hour for running this vision language model on a chosen SageMaker instance. Pricing scales with the instance type you select, so more powerful hardware costs more per hour. Most options run in real-time mode, which serves live requests. One option, ml.g5.48xlarge, also offers batch mode for processing grouped workloads. The remaining choices span several instance families and sizes, letting you match capacity to your inference needs. Billing is usage-based, so you are charged only for the hours each instance runs.
Top-of-mind questions for buyers
One host hour is one hour that a single inference endpoint instance runs. Billing starts when the instance is active and stops when you shut it down. Each instance you run accrues its own hours. The rate differs by the instance type you choose.
Real-time mode serves live requests as they arrive, so the instance stays running to respond instantly. Batch mode processes grouped inputs together, which suits large jobs run at once. Only the ml.g5.48xlarge offers batch mode; all listed instances offer real-time mode.
Software charges apply per hour while the instance runs. A stopped instance does not accrue host-hour software charges. Underlying AWS infrastructure fees may still apply separately depending on your setup. You pay only for the hours each instance is active.
docs.nvidia.com
Helpful?
Vendor refund policy
No Refund
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker model
An Amazon SageMaker model package is a pre-trained machine learning model ready to use without additional training. Use the model package to create a model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Additional details
Inputs
- Summary
Cosmos Reason 2 8B accepts JSON requests via the /invocations API, where the image or video content is provided.
Input Type(s): Text+Video/Image
Input Format(s):
Text: String
Video: mp4
Image: jpg
Input Parameters:
Text: One-dimensional (1D)
Video: Three-dimensional (3D)
Image: Two-dimensional (2D)
Other Properties Related to Input:
Use FPS=4 for input video to match the training setup.
Append Answer the question in the following format: \nyour reasoning\n\n\n\nyour answer\n. in the system prompt to encourage long chain-of-thought reasoning response.
- Input MIME type
- application/json
{
"model": "nvidia/cosmos-reason2-8b",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role":"user",
"content":[
{
"type": "text",
"text": "What is in this image? \nAnswer the question in the following format: <think>\nyour reasoning\n</think>\n\n<answer>\nyour answer\n</answer>."
},
{
"type": "image_url",
"image_url": {
"url": image_data_url
}
}
]
}
],
"temperature": 0.6,
"max_tokens": 200,
}
{
"model": "nvidia/cosmos-reason2-8b",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role":"user",
"content":[
{
"type": "text",
"text": "What is in this image? \nAnswer the question in the following format: <think>\nyour reasoning\n</think>\n\n<answer>\nyour answer\n</answer>."
},
{
"type": "image_url",
"image_url": {
"url": image_data_url
}
}
]
}
],
"temperature": 0.6,
"max_tokens": 200,
}
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
model | The specific model name, e.g., "nvidia/cosmos-reason2-8b". | String | Yes |
messages | Conversation history, typically containing a single "user" message.
| Array of Objects | Yes |
messages[].content | Must contain an object with "type": "image_url or video_url" and an "image_url or video_url" object. The image/video data can be provided via a Base64-encoded data URI (e.g., data:image/png;base64,...) for local files, or as a direct public URL to the online image/video file. | Array of Objects
| Yes |
max_tokens | The maximum number of tokens to generate in the response. | Integer | No |
Resources
Support
Vendor support
Free support via NVIDIA NIM Developer Forum:
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
NVIDIA Cosmos Reason: an open, customizable, 7B-parameter reasoning vision language model (VLM) for physical AI and robotics - enables robots and vision AI agents to reason like humans, using prior knowledge, physics understanding and common sense to understand and act in the real world.
Llama 3.1 Nemotron Nano 8B V1 is a large language model (LLM) which is a derivative of Meta Llama-3.1-8B-Instruct with reasoning capabilities.
NVIDIA Nemotron are open models with open weights, training data, and recipes, delivering leading efficiency and accuracy for building specialized AI agents
Available as a NIM, Llama 3.1 70B-Instruct is a 70B-parameter LLM pretrained, instruction tuned model optimized for multilingual dialogue.
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.