Request for Information (RFI) for Artificial Intelligence (AI) Model Access Services

Quick Summary

The U.S. Department of the Treasury has issued a Request for Information (RFI) for Artificial Intelligence (AI) Model Access Services. This RFI is for market research and acquisition planning, seeking industry feedback to refine the scope and approach for an anticipated acquisition of Frontier AI company tokens. The goal is to establish direct, flexible, and rapid access to evolving AI/Large Language Model (LLM) capabilities. There is no incumbent contractor for this effort. Responses are due by February 23, 2026, at 9:00 AM ET.

Purpose & Scope

Treasury aims to procure sustained, direct access to frontier AI companies through a token-based approach, supporting mission-driven use and adaptability to market advances. This initiative is driven by recent Executive Orders and OMB Memoranda on accelerating Federal AI use. The primary objective is to support software development augmentation, with secondary uses including document analysis, research assistance, and content generation across the Treasury enterprise and other Federal entities.

A "Frontier AI Company" is defined as an organization that develops proprietary LLMs, demonstrates state-of-the-art performance (e.g., MMLU >75%, HumanEval >65%), maintains direct control over model development/deployment/API, and has established commercial API offerings with documented SLAs. Treasury is primarily interested in direct relationships with model developers.

Anticipated Contract Details

• Type: Multiple Indefinite Delivery, Indefinite Quantity (IDIQ) contracts or a similarly flexible vehicle.
• Duration: One base year with up to four option years (totaling up to 5 years).
• Set-Aside: None specified.
• NAICS: 518210 – Data Processing, Hosting, and Related Services
• PSC: DA10 – IT and Telecom – Business Application/Application Development Software as a Service (SaaS)
• Token Framework: Treasury intends to establish a standardized token measurement and pricing framework for consistent cost comparison, budgeting, and transparent usage tracking.
• Refresh: Periodic (potentially annual) refresh of contract holders to incorporate new vendors and capabilities.
• Place of Performance: Services delivered via cloud-based infrastructure from Contractor facilities within the Continental United States, with all data processing and storage also within CONUS.

Key Requirements & Performance Standards

The anticipated acquisition will require:

• Availability: 99.9% uptime during core business hours.
• Response Latency: P95 response time less than 3 seconds.
• Error Rate: Less than 0.1% API error rate.
• Security & Compliance: FedRAMP Moderate authorization (or clear path), data residency/sovereignty controls within CONUS, no government data retention beyond immediate processing, no use of government data for model training without explicit consent. Compliance with EO 14110, OMB M-24-10, NIST AI RMF, NIST SP 800-53, and Section 508.
• Data Handling: Government retains all rights to input and output data.
• Integration: RESTful API access, comprehensive documentation, SDKs, secure authentication, and detailed usage metering.

Anticipated Evaluation Criteria

Future solicitations may evaluate based on:

• Token-Based AI Capability and Governance
• Cost Efficiency and Pricing Transparency
• Security and Compliance
• Management Approach

Response Requirements

Vendors are requested to provide:

• Company information (name, UEI, business size, socio-economic status).
• Existing government contract vehicles for AI tokens.
• Questions (up to 30) and recommendations (up to 25) to refine the requirement.
• Identification of potential barriers to participation.
• Detailed responses (limited to 1 page per question) to 18 specific questions covering: on-premises/cloud considerations, token definition/variability, token usage distribution, government data handling, model availability/versioning, security/authentication, FedRAMP status, consumption mechanisms, service availability/support, pricing structure, exit/transition, challenges with the proposed contractual model, tokenization methodology (including input token counts for representative text samples as per "Table 1"), commercial indemnification, pricing trends, model retirements, capacity management, and enterprise-level policy tooling.
• Marketing materials are not useful and will not be reviewed.

Submission Details

• Due Date: February 23, 2026, 9:00 AM Eastern Time.
• Email: Benjamin.A.Mendelsohn@irs.gov and George.A.Monsivais@irs.gov
• Subject Line: “Treasury Notice – Frontier AI Company Tokens – February 2026”
• Format: Searchable Adobe PDF (version 7.x or earlier), 0.5-inch margins, no smaller than 10-point font.

Important Notes

This is not a solicitation and does not obligate the Government to issue one or award a contract. Responses are for market research and planning purposes only.