Sold by: MongoDB, Inc
Deployed on AWS
Free Trial
Next-generation contextualized chunk embeddings with built-in auto-chunking, overlap support, and extended context length for higher long-document retrieval quality.
Overview
Contextualized chunk embedding models are neural networks that encode both a chunk's own content and the contextual information from the full document into numerical vectors. They are a foundational building block for semantic search/retrieval systems and retrieval-augmented generation (RAG) and directly determine retrieval quality.
voyage-context-4 is the next generation of Voyage AI's contextualized chunk embedding model, designed to deliver higher retrieval accuracy, more effective long-document support, and built-in auto-chunking, with full support for both overlapping and non-overlapping chunks. Users can submit a full document as a single string and let the backend chunk it automatically, eliminating brittle manual chunking and preprocessing logic. The chunked text is returned in the response for inspection and storage.
voyage-context-4 delivers approximately 1.4% NDCG@10 points higher chunk-retrieval quality than voyage-context-3 and approximately 8.4% NDCG@10 points higher than cohere-embed-v4.0 across chunk-level retrieval benchmarks. It removes the 32K-token ceiling that is a challenge for most contextualized chunking models by handling longer documents gracefully through backend partitioning and an extended context window. Enabled by Matryoshka representation learning and quantization-aware training, voyage-context-4 supports embeddings in 2048, 1024, 512, and 256 dimensions, with multiple quantization options.
Learn more about voyage-context-4 here: https://blog.voyageai.com/2026/06/29/voyage-context-4/
Highlights
- Built-in auto-chunking lets users submit a full document as a single string and the backend chunks it automatically, returning chunk text in the response, removing the need to hand-tune chunk sizes and preprocessing logic.
- Improved retrieval quality with robust support for both overlapping and non-overlapping chunks, plus extended context handling via backend partitioning, removing the 32K-token ceiling with voyage-context-3.
- Supports embeddings of 2048, 1024, 512, and 256 dimensions and multiple embedding quantization options, 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).
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 7 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.g7e.2xlarge Inference (Real-Time) Recommended | Model inference on the ml.g7e.2xlarge instance type, real-time mode | $4.49 |
ml.g5.xlarge Inference (Batch) Recommended | Model inference on the ml.g5.xlarge instance type, batch mode | $35.92 |
ml.p5.48xlarge Inference (Real-Time) | Model inference on the ml.p5.48xlarge instance type, real-time mode | $35.92 |
ml.p4de.24xlarge Inference (Real-Time) | Model inference on the ml.p4de.24xlarge instance type, real-time mode | $35.92 |
ml.g7e.4xlarge Inference (Real-Time) | Model inference on the ml.g7e.4xlarge instance type, real-time mode | $4.49 |
ml.g7e.8xlarge Inference (Real-Time) | Model inference on the ml.g7e.8xlarge instance type, real-time mode | $4.49 |
ml.g7e.12xlarge Inference (Real-Time) | Model inference on the ml.g7e.12xlarge instance type, real-time mode | $8.98 |
ml.g7e.24xlarge Inference (Real-Time) | Model inference on the ml.g7e.24xlarge instance type, real-time mode | $17.96 |
ml.g7e.48xlarge Inference (Real-Time) | Model inference on the ml.g7e.48xlarge instance type, real-time mode | $35.92 |
Dimensions summary
You pay an hourly rate for each host you run this embedding model on. Pricing is organized by deployment mode and instance type. Eight dimensions cover real-time inference, where the model runs a persistent endpoint for request-by-request use. These span a range of GPU instance types, from the g7e family in several sizes to the p4de and p5 types. One dimension covers batch inference on a g5.xlarge instance, which runs finite jobs for bulk processing. Your cost scales with how long each instance runs and which instance type you choose.
Top-of-mind questions for buyers
Your total hourly cost combines two parts. Software pricing covers use of the embedding model itself. Infrastructure pricing covers the AWS instance it runs on. Both are billed at an hourly rate. The rates vary by deployment mode, instance type, and region.
Real-time inference runs a persistent endpoint you keep active, so charges accrue for as long as the endpoint runs. Batch inference runs a finite job that processes a dataset and stops. Real-time suits request-by-request use. Batch suits bulk processing of files, where the instance runs only until the job completes.
No. Canceling your subscription does not terminate existing real-time endpoints or batch transform jobs. Running endpoints keep accruing charges until you delete them through the SageMaker console or Studio. To avoid unexpected costs, delete endpoints you no longer need rather than relying on cancellation.
docs.voyageai.com
Helpful?
Vendor refund policy
Refunds are processed according to the EULA. For assistance, contact aws-marketplace@mongodb.com .
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
MongoDB is excited to announce the initial release of voyage-context-4, featuring built-in auto-chunking, overlap support, and extended context handling for higher long-document retrieval quality.
Additional details
Inputs
- Summary
Supply one or more inputs (documents or queries) to vectorise. Use enable_auto_chunking=true to submit full documents for automatic backend chunking with optional overlap, or pass pre-chunked documents as a list of lists.
Note: voyage-context-4 embedding model does NOT support batch transform.
- Limitations for input type
- Max 1,000 inputs per request; max 16,000 total chunks per input; max 120,000 total tokens per request. Note: voyage-context-4 embedding model does NOT support batch transform.
- Input MIME type
- application/json
https://github.com/voyage-ai/voyageai-aws/blob/main/sample_contextualized_embedding_input.json
voyage-context-4 embedding model does NOT support batch transform.
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
inputs | The input texts to be vectorized. Pass a flat list of strings for full documents (with enable_auto_chunking=true) or queries. Pass a list of lists of strings for pre-chunked documents. | Maximum of 1,000 inputs. | Yes |
input_type | The role of the input: query, document, or null. | Default: null | No |
output_dimension | The number of dimensions for resulting embeddings. | One of 2048, 1024, 512, 256. Model default 1024 when null. | No |
output_dtype | Data type for returned embeddings. binary and ubinary return bit-packed 8-bit integers; returned list length is 1/8 of output_dimension. | One of float, int8, uint8, binary, ubinary. Default: float | No |
enable_auto_chunking | When true, the backend automatically chunks each input document. Requires inputs to be a flat list of strings and input_type=document. | Default: false | No |
chunk_size | Target chunk size in tokens when enable_auto_chunking=true. Actual chunk size may be less than the value passed. | Must not exceed 32,000 tokens. Default: 512 server-side. | No |
chunk_overlap | Overlap between consecutive chunks in tokens when enable_auto_chunking=true. Overlapping tokens are billed as input tokens. | Must be smaller than chunk_size. Default: 0 | No |
Resources
Vendor resources
Support
Vendor support
Please email us at aws-marketplace@mongodb.com for inquiries and customer support.
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 reranking model optimized for legal RAG, research, and classification, ranked first on Legal RAG Bench and the Massive Legal Embedding Benchmark (MLEB).
Best-in-class legal enrichment, reranking, and embedding models, optimized for legal knowledge graphs and legal RAG.
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.