1. Definitions
“Adapter” means an object that is designed to be compatible with the designated Model and intended to adjust the Model Output in accordance with the algorithm contained within the Adapter. “Authorized User” means any End User who has been explicitly designated by Customer and granted specific administrative rights and permissions to manage the Token Factory Service. “Customer AI Content” means Customer’s Inputs, Outputs, Adapters, Fine-Tuned Models, Distilled Models, Datasets, Models and smaller Models for Speculative Decoding. For the avoidance of doubt, Customer AI Content is considered Customer Data for purposes of the Agreement (including data deletion, Customer obligations, and indemnification), unless expressly stated otherwise in these Token Factory Terms or an applicable Order Form. “Customer Product” means any product, service, application, or system developed, managed, or distributed by Customer that utilizes or integrates with the Token Factory Service. “Dataset” means any collection of data that is used for training, validating, fine-tuning Models or other purposes, and includes both datasets that are provided by Customer and uploaded to the Platform, as well as those datasets that are created or generated by Customer on the Platform itself. “Distillation” means the process of training a smaller Model to replicate the behavior of a larger Model while maintaining similar performance characteristics. “Distilled Model” means the resulting Model created through Distillation. “Dedicated Endpoint” means configurable deployment of a supported Model with dedicated capacity exclusively for Customer’s use, which Customer may create and manage through the Platform. “End User” means any individual who, subject to compliance with the Agreement (including these Token Factory Terms): (a) accesses or uses the Token Factory Service under Customer’s Account; or (b) uses a Customer Product that utilizes or integrates with the Token Factory Service. “EU AI Act” means Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) as amended. “EULA” means a license agreement, an acceptable use policy, and/or any other similar document published by a Model’s provider and governing the use of the Model plus additional rules applicable to a Model’s usage indicated in Models’ cards. “Fine-Tuned Model” means a Model that has undergone Fine-Tuning. “Fine-Tuning” means the process of adapting Model to enhance its performance on specific tasks or Datasets. “Input” means any data, text, image, audio, video query, or other information Customer submits through the Platform to prompt Models. “Model” means any computational algorithm, system, or framework developed to analyze data and learn patterns for performing specific tasks, including but not limited to prediction, classification, generation, or decision-making. “Operations” means actions performed by Token Factory Service to manage, retrieve, create, or modify cloud resources. “Output” means any text information generated by the Models in response to the Input provided by Customer through the Platform. “Speculative Decoding” means an inference-time method that speeds up Output generation by using a smaller, specific-purpose (“draft”) Model to propose multiple tokens, which are then verified or rejected by a larger (“target”) Model. Speculative decoding reduces the number of steps the target Model must compute while preserving its Output quality and significantly increases the speed of the Services while maintaining the quality of the Model’s responses. “Token Factory API/SDK” means an API, software development kit, or other similar set of protocols and tools that enables End Users to send Input to Models and receive Output. “Token Factory API/SDK Keys” means unique identifiers provided by Nebius to Customer to authenticate and authorize their access to Token API/SDK(s).2. Scope of Use
2.1. Access. Unless otherwise stipulated in an Order Form, subject to Customer’s compliance with the Agreement, including these Token Factory Terms and any applicable EULA or Documentation, Nebius grants Customer the right to access and use the Token Factory Service during the term of the Agreement. 2.2. Inference. The Token Factory Service provide Customer with the inference from available Models via web interface, or Token Factory API/SDK. Customer can set an Input and a set of parameters, which may depend on the applicable Model. In response, Customer receives an Output. Specific functionalities are described in Documentation. 2.3. Dedicated Endpoints. The Token Factory Service provides Customer with Dedicated Endpoints to deploy supported Models with dedicated capacity. Subject to the terms of the Agreement (including these Token Factory Terms and any applicable Documentation), Nebius grants Customer a limited right to create, configure, and manage Dedicated Endpoints through the Platform during the term of the Agreement. Specific functionalities of Dedicated Endpoints are described in Documentation. 2.4. EU AI Act Compliance. Where Customer integrates Token Factory API/SDK into a Customer Product, Customer is the “provider” of the resulting AI system for purposes of Article 3(3) of the EU AI Act and any equivalent or successor legislation. Customer acknowledges and agrees that it is solely responsible for compliance with any obligations applicable to providers of such resulting AI systems under the AI Act. Nothing in these Token Factory Terms, in any documentation, or in any technical feature made available by Nebius shall be construed as Nebius assuming the role of “provider” of any AI system built, deployed, or placed on the market by Customer using Token API/SDK. 2.5. Token Factory API/SDKs. Subject to the terms of the Agreement (and any associated Documentation or Token API/SDK-specific terms made available by Nebius), Nebius grants Customer a limited, revocable, non-exclusive, non-sublicensable, non-transferable right to use the applicable Token API/SDK(s) for the term of the Agreement solely for the purpose of and to the extent reasonably necessary for incorporating such Token API/SDK(s) into a Customer Product in order to make the applicable Services available via such Customer Product (including allowing its End Users to use such Services through such Customer Product). Customer is responsible for maintaining the security of Customer’s Token API/SDK Key and for any activity that occurs while using Customer’s Token API/SDK Key. 2.6. User Administration. Customer shall be responsible for assigning and managing roles of Authorized Users as described on the Platform, including maintaining an up-to-date list of such End Users, and shall promptly revoke or amend associated access rights when no longer required. 2.7. EULAs. Models we provide may be subject to their own EULA, and it is Customer’s responsibility to comply with those EULAs available in Models’ cards, which are hereby incorporated by reference into these Token Factory Terms and constitute a binding part of them. By using the applicable Model, Customer and Customer’s End Users agree to comply with applicable EULA as so incorporated. The use of Customer’s Adapters and smaller Models for Speculative Decoding is restricted to a predefined list of Models specified by Nebius. 2.8. Access Control. Through the settings of the Token Factory Service, Customer can control access to its Models, its smaller Models for Speculative Decoding, and its Adapters. 2.9. Adapters and Speculative Decoding. Customer is solely responsible for (a) ensuring that its Adapters and smaller Models for Speculative Decoding are properly designed and compatible with the specified Model; (b) complying with all applicable laws and regulations in connection with the creation and use of its Models, smaller Models for Speculative Decoding, and Adapters; (c) compliance with each applicable EULA in connection with the creation and use of Customer’s Models, smaller Models for Speculative Decoding, and Adapters. Customer represents and warrants that it has all necessary rights to use its Models, smaller Models for Speculative Decoding, and Adapters, and Customer authorizes Nebius to use them as described in or contemplated under the Agreement. 2.10. Fine-Tuning and Distillation. Fine-Tuned or Distilled Models will only be accessible for use by the Customer who created them. Nebius has a right to use Customer’s Datasets for the purpose of Fine-Tuning or Distillation and a right to use Customer’s Fine-Tuned or Distilled Model for the purpose of making it available to Customer via the Services. The use of Fine-Tuned or Distilled Models via the Services is subject to the following: (a) Nebius reserves the right to impose a limit on the number of Fine-Tuned or Distilled Models that can be used by Customer at any given time; (b) Nebius has a right to delete Customer’s Fine-Tuned or Distilled Model from the Services upon providing Customer with at least three (3) days’ prior written notice, or immediately if required by applicable law, security concerns, or risk of harm to the Services; and (c) Fine-Tuning and Distillation are restricted to a predefined list of Models specified by Nebius. Customer is solely responsible for (i) ensuring that the Datasets Customer uploads comply with all applicable laws, regulations and the Agreement; and (ii) compliance with respective EULA in connection with the creation and use of Fine-Tuned or Distilled Models. Customer represents and warrants that Customer has all necessary rights to use training and validation Datasets and Fine-Tuned or Distilled Models and authorize Nebius to use them as described herein. 2.11. Professional Services. Nebius may, at Customer’s request, provide professional services related to the optimization, deployment, and management of Models and associated infrastructure (“Professional Services”). These Professional Services may include but are not limited to: model-to-GPU optimization, model-level optimization, API-level optimization, model training services (such as fine-tuning, distillation, and training of custom models to improve latency and quality) and pipeline quality improvement (including prompt engineering, task decomposition, retrieval pipeline optimization, and retrieval-augmented generation). Professional Services may be provided upon written request by Customer to Nebius’ designated email address for Professional Services, as specified by Nebius from time to time, or by execution of an Order Form. All Models and Adapters created or modified for Customer in the course of Professional Services shall be deemed Customer’s Models and Adapters. If the Professional Services result in the creation or modification of a General Purpose AI (GPAI) model, as defined under the EU AI Act, Customer shall be deemed the provider of this GPAI model and shall bear all corresponding obligations under applicable law, including but not limited to the EU AI Act. Subject to the terms of the DPA, Nebius has a right to use Customer’s Datasets (provided hereunder) for the purpose of providing Professional Services. Customer is solely responsible for (a) ensuring that the training and validation Datasets Customer uploads comply with all applicable laws, regulations and the Agreement; and (b) compliance with respective EULA in connection with the creation and use of modified Models. Customer represents and warrants that Customer has all necessary rights to use Customer’s Datasets and Fine-Tuned or Distilled Models and to authorize Nebius to use them as described herein.3. Rights and Ownership
3.1. Customer Content. As between the parties and to the extent permitted by law, Customer retains ownership of Customer Content, including its Input and Output. Customer acknowledges that Outputs may not be unique, and other users of the Token Factory Service may receive the same or similar Outputs. Customer is solely responsible for obtaining all necessary rights, approvals, and consents to provide Input and for the processing of Input by the Token Factory Service. For clarity, nothing in these Token Factory Terms limits Nebius’s rights in Feedback as defined in the Agreement. Customer acknowledges and agrees that the use of any third-party Model through the Token Factory Service does not create, and shall not be construed as creating, any joint development, joint ownership, implied license, or other proprietary interest in such Model or in the related materials of the Model’s provider. All right, title, and interest in and to such Models and related materials remain exclusively with their respective providers. 3.2. Certain Models. In addition to its other intellectual property rights under the Agreement, Nebius retains ownership of any smaller Models for Speculative Decoding created by or for Nebius.4. Restrictions
In addition to the other restrictions set forth in the Agreement, Customer will not:- Use any Output or other results of the Token Factory Service to develop any service that competes with the Services;
- Attempt to bypass, disable, or circumvent any safety filters, content moderation, rate limits, or other usage controls applicable to any Model or the Token Factory Service;
- Represent any Output as human-generated where disclosure of AI involvement is required by applicable law or would be materially deceptive;
- Use the Token Factory Service for fully automated decision-making that produces legal or similarly significant effects on individuals without appropriate human oversight;
- Use any Token Factory Service as or in a high-risk AI system or in prohibited AI practices as defined in the EU AI Act; or
- Violate any applicable EULA.
5. Data Use and Privacy
5.1. Processing of Input and Output. Nebius processes Input to generate Output and to operate, maintain, and provide the Token Factory Service. Subject to the terms of the DPA, Nebius may access, use, host, cache, store, copy, modify, and otherwise process Input and Output to (a) provide, maintain, monitor, protect, and support the Services, Token Factory Service or (b) train smaller Models used for Speculative Decoding. These rights and licenses are royalty free, transferable, sub-licensable, and worldwide. Subject to the terms of the DPA, Nebius may collect usage data and technical logs relating to Customer’s use of the Token Factory Service for security, support, and billing. Customer shall submit the Input in compliance with all applicable laws, rules and regulations, including, without limitation, Applicable Data Protection Laws, and Customer shall obtain such consents or other legal permissions as are required by Applicable Data Protection Laws and the DPA. 5.2. Processing of Personal Data. In connection with Customer’s use of the Token Factory Service, subject to the terms of the DPA, Nebius processes certain technical metadata that may constitute personal data under Applicable Data Protection Laws. Such metadata may include log data and IP addresses. This processing may be reasonably necessary to ensure the proper functioning, security, and integrity of the Token Factory Service or the Platform. 5.3. Deletion of Data. Subject to the terms of the DPA, Nebius reserves the right, and has absolute discretion, to remove, screen, or delete any Inputs or Outputs at any time, for any reason, and without notice. 5.4. Opt Out. If Customer prefers that Nebius (i) does not store Customer’s Inputs and Outputs for the purposes of providing the Services, or (ii) does not collect and process them for Speculative Decoding, Customer may opt out at any time. Customer can indicate such preference through the opt-out mechanism available in the onboarding form or by submitting a request to Nebius’ support team via email at tokenfactory-support@nebius.com. Please be aware that opting out may, in some cases, limit the Services’ ability to address Customer’s specific use case effectively. Customer’s opt-out preferences under this Section 5.4 will be implemented in a manner consistent with the Agreement and the DPA, and may affect the storage and processing of Customer AI Content under applicable Service-specific deletion rules. 5.5. Further Information. For more details about how Nebius processes Personal Data, categories of recipients, data subject rights, and relevant contact information, please refer to Nebius’ Privacy Policy, available at: https://docs.nebius.com/legal/privacy.6. Payments
6.1. Fees. Unless otherwise set forth in the applicable Order Form, Token Feature-related fees are on a pre-paid basis, and the amounts of such fees are stated on or via the Platform. Customer is required to deposit funds into the Customer Account in accordance with Nebius’ or the Platform’s instructions. Any deposit not used within one year of when made will automatically expire (without any refund) subject to any different refund or expiry terms expressly set forth in an applicable Order Form. 6.2. Credits. If Customer is granted or otherwise receives any free credits for use of the Services, Nebius reserves the right to modify or revoke any such credits at any time and without prior notice.7. Disclaimers
7.1. General Disclaimer. In addition to the other disclaimers set forth in the Agreement, Customer acknowledges and agrees that (a) Output and Operations may be inaccurate, incomplete, inappropriate, invalid, unreliable, insecure, or offensive; (b) Customer will carefully evaluate all Output and Operations before relying on or using them; (c) Customer is responsible for ensuring that any Output incorporated into Customer’s products, services, or materials complies with applicable laws and does not infringe third-party intellectual property rights or other rights; (d) Customer is solely responsible for all Operations it authorizes and any consequences thereof; (e) Nebius does not guarantee that Customer’s smaller Models for Speculative Decoding and Adapters will function as intended or be compatible with the designated Model; (f) Nebius reserves the right to impose a limit on the number of Customer’s Models, smaller Models for Speculative Decoding, and Adapters that can be used by Customer at any given time; (g) Nebius has a right to delete Customer’s Models, smaller Models for Speculative Decoding, and Adapters from the Services upon providing Customer with at least three (3) days’ prior written notice, or immediately if required by applicable law, security concerns, or risk of harm to the Services; (h) Nebius is not responsible for ensuring the functionality or performance of Customer’s Models, smaller Models for Speculative Decoding, or Adapters, including but not limited to compatibility with the Services, or Outputs produced by them; (i) the Token Factory Service are provided on an “as-is” and “as-available” basis and at Customer’s sole risk; and (k) Nebius can suspend or terminate Customer’s access to a specific Model upon a written request from such Model’s provider as a result of a violation of the applicable EULA. 7.2. Additional Output Disclaimer. Due to the nature of the Token Factory Service, Nebius does not represent or warrant: (a) that Output does not incorporate or reflect third-party content or materials, or (b) that Output will not infringe third-party intellectual property rights. Customer assumes all responsibility for its use of the Token Factory Service and Output, and Nebius shall not be liable for any damages arising from or related to such use. 7.3. Additional IP Disclaimer. Claims of intellectual property infringement or misappropriation by Output are not included in the covered third-party claims for any indemnification by Nebius under the Agreement. For clarity, this Section 7.3 supplements, and does not replace, the indemnity exclusions set forth in the Agreement. 7.4. No Service Level Agreement. The Service Level Agreement (“SLA”) available at https://docs.nebius.com/legal/sla does not apply to the Token Factory Service. Nebius does not provide any service level commitments, uptime guarantees, or associated compensation for the Token Factory Service. For the avoidance of doubt, no compensation or credits shall be available to Customer in connection with any unavailability or underperformance of the Token Factory Service, unless agreed otherwise in writing.8. Indemnification
8.1. Additional Indemnification. In addition to Customer’s other indemnification obligations under the Agreement, Customer will defend (using counsel reasonably acceptable to Nebius), hold harmless and indemnify Nebius, its affiliates, agents, subcontractors, partners, licensors and each of their respective employees, officers, directors from any and all losses, damages, judgments, liabilities, costs or expenses, including without limitation attorneys’ fees and legal expenses, arising out of or relating to any third party (including governmental authority) claims, causes of action, investigations, or proceedings concerning any Customer AI Content (or any access, use, disclosure, commercialization or other exploitation of or reliance upon any Customer AI Content).9. Miscellaneous
9.1. Modifications. Nebius may update these Token Factory Terms from time to time. Updates will be communicated consistent with the notice provisions of the Agreement and will become effective as stated in such notice. Continued use of the Token Factory Service constitutes acceptance of the updated Token Factory Terms. 9.2. Conflict. In the event of any conflict between these Token Factory Terms and other components of the Agreement, the hierarchy set forth in the Agreement (including the precedence of Service-Specific Terms for service-specific subject matter) shall apply. Otherwise, the Agreement remains in full force and effect.Web address: https://docs.nebius.com/legal/token-factory Publication date: September 15, 2026 Effective date: September 15, 2026