Overview
Zuse is Bineric's decision model for European companies. Give it a document, such as an email, a ticket, a claim or a contract clause, and the questions you want answered. For each question it picks from the options you set and returns how sure it is, so you can act automatically on the confident answers and send the rest to a person. It is built for routing, triage, compliance checks and filling fixed fields, where a chat model is slower, costlier and harder to audit. It cannot answer outside the options you give it.
Zuse uses the same request format as TypeSafe's Jev, so teams that know Jev can start straight away, with the model running inside their own AWS account instead of a hosted API. Requests in the OpenAI Decisions format are accepted too.
Zuse 9B is the fast, low-cost model in the family: it fits the smallest GPU instances and answers about twice as fast as Zuse 27B, at 3.8 points lower accuracy. It is built on Qwen3.5-9B-Base and learned from Zuse 27B's own answers; Zuse 27B was trained by Bineric on permissively licensed public data and data we generated and translated ourselves. No customer data was used, and no proprietary model's outputs. Bineric AS, Oslo, Norway.
Languages: 83 languages score 85% or more on reading and topic tests (Belebele and SIB-200, 210 languages tested).
Trained on 27 European languages: Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Icelandic, Italian, Latvian, Lithuanian, Maltese, Norwegian Bokmal, Norwegian Nynorsk, Polish, Portuguese, Romanian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.
17 more European languages: Albanian, Asturian, Basque, Belarusian, Bosnian, Catalan, Galician, Irish, Limburgish, Luxembourgish, Macedonian, Occitan, Russian, Serbian, Sicilian, Venetian, Welsh.
39 languages beyond Europe: Afrikaans, Arabic, Awadhi, Azerbaijani, Banjar, Bhojpuri, Cantonese, Chhattisgarhi, Chinese (Simplified), Chinese (Traditional), Dari, Egyptian Arabic, Esperanto, Hebrew, Hindi, Indonesian, Japanese, Kannada, Korean, Kyrgyz, Magahi, Maithili, Malay, Marathi, Minangkabau, Papiamento, Sanskrit, Sinhala, South Levantine Arabic, Ta'izzi-Adeni Arabic, Tagalog, Tamil, Tatar, Tok Pisin, Tunisian Arabic, Turkmen, Uyghur, Vietnamese, Yiddish.
Highlights
- One pass answers many questions about a document: yes/no, a choice among up to 255 options, or a position on a scale. Each answer has a calibrated probability, so you automate the confident cases and route the rest to people.
- 83 languages at 85% or more, with 27 European languages trained on. 83.6% accuracy with these 4-bit weights on 20,044 questions from 11 task families it never trained on (Zuse 27B: 87.4%). Documents up to 64k tokens (tested to 60k on ml.g5.xlarge).
- Runs inside your AWS account on one 24 GB GPU, down to the smallest (ml.g5.xlarge), with network isolation, so your documents never leave your VPC. Short requests take under a second. Same request format as TypeSafe's Jev.
Details
Features and programs
Financing for AWS Marketplace purchases
Pricing
Free trial
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.g5.xlarge Inference (Real-Time) Recommended | Model inference on the ml.g5.xlarge instance type, real-time mode | $0.50 |
ml.g5.2xlarge Inference (Batch) Recommended | Model inference on the ml.g5.2xlarge instance type, batch mode | $0.50 |
ml.g6.xlarge Inference (Real-Time) | Model inference on the ml.g6.xlarge instance type, real-time mode | $0.50 |
ml.g5.2xlarge Inference (Real-Time) | Model inference on the ml.g5.2xlarge instance type, real-time mode | $0.50 |
ml.g5.4xlarge Inference (Real-Time) | Model inference on the ml.g5.4xlarge instance type, real-time mode | $0.50 |
ml.g6.2xlarge Inference (Real-Time) | Model inference on the ml.g6.2xlarge instance type, real-time mode | $0.50 |
ml.g6.4xlarge Inference (Real-Time) | Model inference on the ml.g6.4xlarge instance type, real-time mode | $0.50 |
ml.g6.2xlarge Inference (Batch) | Model inference on the ml.g6.2xlarge instance type, batch mode | $0.50 |
ml.g6e.xlarge Inference (Real-Time) | Model inference on the ml.g6e.xlarge instance type, real-time mode | $0.75 |
ml.g6e.2xlarge Inference (Real-Time) | Model inference on the ml.g6e.2xlarge instance type, real-time mode | $0.75 |
Dimensions summary
Top-of-mind questions for buyers
Vendor refund policy
No refunds on hourly usage. For billing questions or problems, email support@bineric.com and we will answer within one business day (CET).
How can we make this page better?
Legal
Vendor terms and conditions
Content disclaimer
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.
Version release notes
Long documents read in chunks: tested to 60k tokens on ml.g5.xlarge, 30 s. Bounded request queue, so a burst gets a quick retry message instead of a timeout. Same model and same accuracy as 1.0.
Additional details
Inputs
- Summary
JSON: a document ('state', text up to 64k tokens) and named typed questions ('noul', 'choice', 'score'). Batch transform: JSON Lines, one request per line.
- Limitations for input type
- Text only, 27 European languages; up to 64k tokens per document; up to 255 options per choice question; 2-10 levels per score question.
- Input MIME type
- application/json, application/jsonlines
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
state | The document to decide on (string or JSON). | - | Yes |
questions | Map of question name to {type, instructions, criteria}. | - | Yes |
Resources
Vendor resources
Support
Vendor support
Email support@bineric.com . We answer within one business day (CET). Usage guide, sample requests and a notebook:
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.