Sold by: Widn.AI
Deployed on AWS
xCOMET provides explainable quality estimation for machine translations, offering detailed error detection alongside quality scores.
Overview
xCOMET is an explainable quality estimation model based on Unbabel's powerful COMET framework, optimized for transparent translation quality assessment. This is a commercial-friendly offering of the model publicly available at https://huggingface.co/Unbabel/XCOMET-XL . Beyond providing quality scores, it identifies specific errors in translations by detecting error spans. This ~3.5B parameter model evaluates machine translations by comparing source and translated texts, and optionally reference translations, providing detailed feedback that aligns with the MQM (Multidimensional Quality Metrics) framework. This service deployment provides a commercial-friendly way to utilize the model's capabilities.
Highlights
- **xCOMET** excels at explainable quality estimation tasks: * Reference-free and reference-based translation evaluation (QE and MT evaluation) * Detailed error detection through error spans * Real-time quality evaluation of translations * Support for diverse language pairs * Integration with translation workflows
- **xCOMET** was trained on a diverse multilingual dataset comprising millions of high-quality human judgments across various domains. The model provides scores between 0 and 1 (where higher is better) along with error spans that help identify and understand specific translation errors. While it excels in many languages, performance may vary for low-resource languages or highly specialized technical content.
- **xCOMET** supports quality estimation across a wide range of languages including major languages like English, Chinese, Spanish, French, German, Japanese, Korean, and Arabic, as well as less common languages such as Welsh, Malagasy, and Kurdish. The model can evaluate translation quality between any language pair from its supported set of over 90 languages, making it versatile for global translation needs.
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
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 | $40.00 |
ml.g5.2xlarge Inference (Real-Time) Recommended | Model inference on the ml.g5.2xlarge instance type, real-time mode | $40.00 |
ml.m5.2xlarge Inference (Batch) | Model inference on the ml.m5.2xlarge instance type, batch mode | $40.00 |
ml.m5.2xlarge Inference (Real-Time) | Model inference on the ml.m5.2xlarge instance type, real-time mode | $40.00 |
ml.m5.xlarge Inference (Real-Time) | Model inference on the ml.m5.xlarge instance type, real-time mode | $40.00 |
Dimensions summary
You pay by the hour for running this translation evaluation model, billed per host hour. Pricing splits across two factors. First, the instance type: ml.m5.xlarge and ml.m5.2xlarge are CPU-based, while ml.g5.2xlarge adds GPU resources for more demanding workloads. Second, the inference mode: batch mode processes grouped inputs together, while real-time mode handles live requests. The ml.m5.2xlarge and ml.g5.2xlarge options support both modes, while ml.m5.xlarge is offered in real-time mode only. You choose the combination that fits your workload, and charges accrue based on hours each instance runs.
Top-of-mind questions for buyers
A host hour is one hour that a single inference instance runs. You pay for each hour the instance stays active, regardless of how many inputs it processes. Billing accrues while the instance is running and stops when you shut it down.
The ml.m5.xlarge and ml.m5.2xlarge options use CPU resources. The ml.g5.2xlarge option adds GPU resources for heavier workloads. The model has about 3.5 billion parameters, so GPU-backed instances suit demanding or faster inference. CPU instances fit lighter evaluation tasks.
Both modes bill per host hour the instance runs. Batch mode processes grouped inputs together, so you run the instance only during job windows. Real-time mode keeps an instance available for live requests, so hours accrue whenever that endpoint stays active.
huggingface.co
Helpful?
Vendor refund policy
No refunds. Please contact support+aws@widn.ai 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
Fix issue with loading tokenizer.
Additional details
Inputs
- Summary
The model accepts JSON input containing batches of source-translation pairs with optional reference translations. See the notebook for examples of inputs.
- Limitations for input type
- The input must be properly formatted JSON with required fields. Each batch can contain multiple translation pairs for efficient processing.
- Input MIME type
- application/json
https://github.com/widn-ai/aws-marketplace
https://github.com/widn-ai/aws-marketplace
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
data[] | An array of objects, each containing source (`src`), translation (`mt`), and optionally reference (`ref`) text. | Type: FreeText | Yes |
batch_size | Number of translations to process in one batch. | Default value: 8
Type: Integer
Minimum: 1
Maximum: 64 | No |
Resources
Vendor resources
Support
Vendor support
Contact us at support+aws@widn.ai
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
Comet's machine learning platform integrates with your existing infrastructure and tools so you can reproduce, debug, manage, visualize, and optimize model - from training runs to production monitoring. Add two lines of code to your notebook or script and automatically start tracking code, hyperparameters, metrics, and more, so you can compare and reproduce training runs.
The official foundation model for image generation from Stability. Deploy this optimized instance and serve generative AI within minutes.
Real-time translation in +50 languages for Zoom, Google Meet, and Teams - with just one click.
A fully ready-to-use Fooocus server for AI-powered image generation with pre-configuration for deployment on AWS. Web interface and HTTPS access. Simply launch the AMI and open the image generation interface in your web browser via the server's IP address. Fooocus is a simplified and more user-friendly alternative to Stable Diffusion WebUI (Automatic1111).
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.