Sold by: MongoDB, Inc
Deployed on AWS
Free Trial
State-of-the-art text embedding model with the best general-purpose and multilingual retrieval quality. 32K context length.
Overview
Text embedding models are neural networks that transform texts into numerical vectors. They are a crucial building block for semantic search/retrieval systems and retrieval-augmented generation (RAG) and are responsible for the retrieval quality. voyage-3-large is a state-of-the-art general-purpose and multilingual embedding model that ranks first across eight evaluated domains spanning 100 datasets, including law, finance, and code. It outperforms OpenAI-v3-large and Cohere-v3-English by an average of 9.74% and 20.71%, respectively. Enabled by Matryoshka learning and quantization-aware training, voyage-3-large supports smaller dimensions and int8 and binary quantization that dramatically reduce vectorDB costs with minimal impact on retrieval quality. Latency is 90 ms for a single query with at most 100 tokens, and throughput is 12.6M tokens per hour at $0.22 per 1M tokens on an ml.g6.xlarge. Learn more about voyage-3-large here: https://blog.voyageai.com/2025/01/07/voyage-3-large/
Highlights
- Outperforms OpenAI-v3-large and Cohere-v3-English by an average of 9.74% and 20.71%, respectively, across 100 datasets, spanning eight diverse domains, including law, finance, and code.
- Supports embeddings of 2048, 1024, 512, and 256 dimensions and offers multiple embedding quantization, including float (32-bit floating point), int8 (8-bit signed integer), uint8 (8-bit unsigned integer), binary (bit-packed int8), and ubinary (bit-packed uint8).
- 32K token context length. Latency is 90 ms for a single query with at most 100 tokens, and throughput is 12.6M tokens per hour at $0.22 per 1M tokens on an ml.g6.xlarge.
Details
Sold by
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
Free trial
Try this product free for 5 days according to the free trial terms set by the vendor.
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.g5.2xlarge Inference (Batch) Recommended | Model inference on the ml.g5.2xlarge instance type, batch mode | $2.27 |
ml.g6.xlarge Inference (Real-Time) Recommended | Model inference on the ml.g6.xlarge instance type, real-time mode | $1.69 |
ml.g5.2xlarge Inference (Real-Time) | Model inference on the ml.g5.2xlarge instance type, real-time mode | $2.27 |
ml.g5.4xlarge Inference (Real-Time) | Model inference on the ml.g5.4xlarge instance type, real-time mode | $3.05 |
ml.g5.8xlarge Inference (Real-Time) | Model inference on the ml.g5.8xlarge instance type, real-time mode | $4.59 |
ml.g5.xlarge Inference (Real-Time) | Model inference on the ml.g5.xlarge instance type, real-time mode | $2.11 |
ml.g6.2xlarge Inference (Real-Time) | Model inference on the ml.g6.2xlarge instance type, real-time mode | $1.83 |
ml.g6.4xlarge Inference (Real-Time) | Model inference on the ml.g6.4xlarge instance type, real-time mode | $2.48 |
ml.g6.8xlarge Inference (Real-Time) | Model inference on the ml.g6.8xlarge instance type, real-time mode | $3.78 |
Dimensions summary
You pay by the hour for the compute instance that runs this embedding model. Charges scale with the instance you select and the inference mode you choose. Real-time mode is available across four ml.g5 sizes (xlarge, 2xlarge, 4xlarge, 8xlarge) and four ml.g6 sizes (xlarge, 2xlarge, 4xlarge, 8xlarge). Larger instance sizes carry more compute and cost more per host hour. Batch mode is offered on the ml.g5.2xlarge instance only. Billing is usage-based, tied to host hours, so you pay for the time each instance runs.
Top-of-mind questions for buyers
One host hour is one hour that a single inference instance runs. This model turns text into vector embeddings, numeric arrays that capture meaning. You pay for each hour the instance stays active, whether serving real-time queries or processing batch jobs. Counting is based on running time, not requests.
Both meter by host hour on the instance you run. Real-time mode keeps an instance active to answer queries as they arrive, so you pay for uptime. Batch mode processes grouped inputs and is offered only on the ml.g5.2xlarge instance. Real-time suits live query workloads; batch suits scheduled bulk jobs.
Charges accrue for every hour an instance runs, even when it handles no queries. Real-time mode keeps the instance active and billable until you stop it. To avoid charges on idle capacity, stop instances you are not using. Stopped instances may still incur underlying AWS storage fees.
www.mongodb.com
Helpful?
Vendor refund policy
Refunds to be processed under the conditions specified in EULA. Please contact aws-marketplace@mongodb.com for further assistance.
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:
Version release notes
We are excited to announce the initial release of voyage-3-large.
Additional details
Inputs
- Summary
- input: str or List[str] - Text(s)
- input_type: str, optional (default=null) - "query" or "document".
- truncation: bool, optional (default=True) - Truncate input.
- output_dimension: int, optional (default=null) - Dimensions for embeddings.
- output_dtype: str, optional (default="float") - Embedding data type
- encoding_format: str, optional (default=null) - Encoding (e.g., Base64)
- id: str, optional (default=null) - Batch request ID.
- Limitations for input type
- The maximum tokens for each text is 32K, the maximum length of the list is 128, and the total number of tokens in the list is at most 32K.
- Input MIME type
- text/csv, application/json, application/jsonlines
https://github.com/voyage-ai/voyageai-aws/blob/main/sample_embedding_input.json
https://github.com/voyage-ai/voyageai-aws/blob/main/sample_batch_input_embedding.jsonl
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
input | A single text string, or a list of texts as a list of strings. | Type: FreeText Limitations: The maximum tokens for each text is 32K, the maximum length of the list is 128, and the total number of tokens in the list is at most 32K. | Yes |
input_type | Type of the input text. Default to null. Other options: "query", "document". | Default value: null Type: FreeText | No |
truncation | Whether to truncate the input texts to fit within the context length. - If True, over-length input texts will be truncated to fit within the context length. - If False, an error will be raised if any given text exceeds the context length. | Default value: True Type: Categorical Allowed values: True, False | No |
output_dimension | The number of dimensions for resulting output embeddings. Options: 2048, 1024, 512, 256, `null`. If `null`, then output dimension is 1024. | Default value: null Type: Integer | No |
output_dtype | The data type for the resulting output embeddings. * "float": 32-bit single-precision floating-point numbers. This is provides the highest precision / retrieval accuracy. * "int8" and "uint8": 8-bit integers ranging from -128 to 127 and 0 to 255, respectively. * "binary" and "ubinary": 8-bit integers that represent bit-packed, quantized single-bit embedding values: int8 for binary and uint8 for ubinary. The length of the returned list of integers is 1/8 of output_dimension. | Default value: float Type: Categorical Allowed values: float, int8, uint8, binary, ubinary | No |
encoding_format | Format in which the embeddings are encoded. * None (default): embeddings are represented as a list of numbers in the data type specified by the output_dtype parameter (default is float); * "base64": embeddings are Base64-encoded NumPy array of (1) numpy.float32 for output_dtype=float, (2) numpy.int8 for output_dtype=int8, binary, and (3) numpy.uint8 for output_dtype=uint8, ubinary | Default value: null Type: Categorical Allowed values: base64, null
| No |
id | Batch transform request ID. If specified, this will be returned in the output.
| Default value: null Type: FreeText | No |
Support
Vendor support
Please email us at contact@voyageai.com for inquiries and customer support. Join our Discord channel: https://discord.gg/zAU7GQEmvT https://docs.voyageai.com/docs/aws-marketplace-model-package .
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
A state-of-the-art legal embedding model optimized for semantic search, RAG, and document classification, ranked first on the Massive Legal Embedding Benchmark.
Contextualized chunk embeddings optimized for general-purpose and multilingual retrieval quality. 32K context length.
Text embedding model optimized for general-purpose (including multilingual) retrieval/search and AI applications. 32K context length.
Text embedding model optimized for general-purpose retrieval quality, latency, and cost for AI applications. 32K context length.
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.