Turn meeting notes, project updates and vendor correspondence into draft summaries and action candidates in English, Spanish and Portuguese. Each item includes the complete selected source segment and Unicode offsets so a reviewer can return to the original statement.
The language model selects source segments; the server copies their text. This verifies literal source location, not semantic entailment. Every response requires human review. Action owners are explicitly named individuals from the source, and deadlines are valid literal ISO dates; unknown values remain null.
Run the dedicated GPU container in your own AWS account with inference-container network isolation. The weights and serving dependencies are included in the image. The product uses the official pinned Qwen/Qwen3.8-27B-FP8 checkpoint without additional training. It accepts bounded text, processes one structured record per invocation, and supports JSON and line-split JSON Lines batch input. Oversized inputs are rejected without truncation.
The product does not execute tools or assign tasks. It provides no OCR, audio transcription or image processing. Summary factuality and independent native-language acceptance remain unmeasured; review every result before use.
Watch the English walkthrough and the vertical Short. The illustrative interface replays unchanged actual certified SageMaker outputs for fictional English, Spanish and Portuguese inputs. It does not start a new inference job and is not included in the model subscription. AI-generated English narration uses NVIDIA Magpie on Jetson AGX Thor. Every draft requires human review.
Highlights
Summaries and action candidates include a source segment and exact source text
Owner and deadline remain null unless the source states a named person and ISO date
Bounded JSON and JSON Lines requests; oversized inputs are rejected without truncation
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.
You pay by the host hour for model inference, so cost scales with how long each instance runs. Two options let you match billing to your workload. The ml.g6e.4xlarge instance runs in real-time mode, for inference that responds on demand. The ml.g6e.2xlarge instance runs in batch mode, for processing grouped requests. These are independent instance-and-mode choices, not stacked tiers. The instance type and mode together set the hourly rate. AWS infrastructure, data transfer, and any third-party model-provider charges bill separately from this software usage.
Top-of-mind questions for buyers
What does one host hour measure, and am I charged when the instance is stopped?
A host hour meters time the inference instance runs. Billing tracks the hours each instance stays active. A fully stopped instance stops accruing software charges. Stopped instances may still incur underlying AWS storage costs. The software license meters running time only.
What is the difference between real-time mode and batch mode for inference billing?
Both bill per host hour. The ml.g6e.4xlarge in real-time mode handles requests that respond on demand, running while you need live responses. The ml.g6e.2xlarge in batch mode processes grouped requests together. Choose real-time for interactive workloads and batch for scheduled, grouped jobs.
What costs are separate from these inference host-hour charges?
AWS infrastructure such as EC2, storage, networking, and data transfer bills separately. Any third-party model-provider charges are also separate. The host-hour rate covers the software usage only. Review each AWS Marketplace listing for current infrastructure pricing before you subscribe.
softwaresushi.com
Helpful?
Vendor refund policy
Software charges stop when the endpoint is deleted or a batch job ends. If the product did not work as described, email info@softwaresushi.com within 30 days with your AWS account ID and endpoint or job name. Software Sushi will refund the software charges for that use. AWS infrastructure charges are separate and are not covered.
How can we make this page better?
Tell us how we can improve this page, or report an issue with this product.
Give us feedbackReport a problem with this product or seller
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.
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:
Real-time inference
Deploy the model as an API endpoint for your applications. When you send data to the endpoint, SageMaker processes it and returns results by API response. The endpoint runs continuously until you delete it. You're billed for software and SageMaker infrastructure costs while the endpoint runs. AWS Marketplace models don't support Amazon SageMaker Asynchronous Inference. For more information, see Deploy models for real-time inference .
Batch transform
Deploy the model to process batches of data stored in Amazon Simple Storage Service (Amazon S3). SageMaker runs the job, processes your data, and returns results to Amazon S3. When complete, SageMaker stops the model. You're billed for software and SageMaker infrastructure costs only during the batch job. Duration depends on your model, instance type, and dataset size. AWS Marketplace models don't support Amazon SageMaker Asynchronous Inference. For more information, see Batch transform for inference with Amazon SageMaker AI .
Version release notes
0.1.0 drafts review-required summaries and action candidates in ES/PT/EN with exact selected source segments and Unicode offsets. Network-isolated runtime and actual batch certification passed on the pinned image. Development-only action metadata results: EN 19/20, ES 20/20, PT 19/20 exact; all 60 admitted. This does not measure customer accuracy or summary factuality. Human review remains required for every result.
Additional details
Inputs
Outputs
Usage instructions
Sample notebooks
Inputs
Summary
One UTF-8 JSON request per invocation with language and segments. JSON Lines uses one structured request per line for line-split batch transform.
Limitations for input type
language must be en, es, or pt. Include 1 to 64 segments; each segment has a unique id and nonempty text. Text is at most 2,000 characters per segment and the sum of all segment-id and text characters is at most 16,000. HTTP body limit is 96 KiB. The actual model input must fit 3,072 tokens and output is limited to 1,024 tokens. Oversized input is rejected; no silent truncation. One structured record is processed per invocation.
Input MIME type
application/json, application/jsonlines
Real-time inference sample input data
{"language": "en", "segments": [{"id": "s1", "text": "Maya Chen will send the signed transition plan by 2027-06-18."}]}
Batch transform sample input data
{"language": "en", "segments": [{"id": "s1", "text": "Maya Chen will send the signed transition plan by 2027-06-18."}]}
{"language": "es", "segments": [{"id": "s1", "text": "Luc\u00eda Mena enviar\u00e1 el cronograma firmado antes del 2027-06-18."}]}
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name
Description
Constraints
Required
language
Requested output language.
One of en, es, or pt.
Yes
segments
Ordered business-source text segments to summarize.
1 to 64 unique segment ids; nonempty text, at most 2,000 characters each and 16,000 combined id-plus-text characters.
Contact Software Sushi at info@softwaresushi.com with the model version, AWS region, instance type, and request ID. Do not send source documents, credentials, or private customer data. No response-time SLA is advertised.
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.
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.