Find and win work
Stay compliant
Research and case law
Agents and integrations
Read
Regulations and policy
Contracts and spending
Audit and compliance
The GSA Multiple Award Schedule is how agencies buy from a contract a company already holds. GovConDash reads it SIN by SIN: what each Special Item Number covers, every contract GSA eLibrary lists under it, and the orders agencies have placed on each.
And the schedule just changed. MAS Refresh #34, posted November 12, 2026, brings the GSAR overhaul to schedule contracts, adds a new clause for AI and large language model offerings, and updates the facilities and cybersecurity SINs.
Special Item Numbers
REFRESH 34 WORKBOOK
Schedule contracts
LISTED IN GSA ELIBRARY
Small business
AS ELIBRARY MARKS THEM
Obligated on orders
AS OF NOVEMBER 22, 2026
MAS is one solicitation, 47QSMD20R0001, split into 12 large categories, 79 subcategories and 277 Special Item Numbers. A contractor proposes the SINs it can support and adds more later by modification.
GSA updates the solicitation through numbered refreshes and brings existing contracts along by mass modification, so the schedule tracks regulation changes without a recompete.
63 of the 277 SINs are marked in GSA's workbook for Cooperative Purchasing or Disaster Purchasing, which lets state and local governments order from them.
277 SINs cover services or services with products, from IT and professional services to facilities, logistics and human capital. Each states its NAICS codes, PSC and maximum order.
34 SINs in 10 subcategories
23 SINs in 6 subcategories
22 SINs in 9 subcategories
14 SINs in 4 subcategories
40 SINs in 9 subcategories
21 SINs in 9 subcategories
11 SINs in 3 subcategories
42 SINs in 10 subcategories
15 SINs in 6 subcategories
27 SINs in 6 subcategories
22 SINs in 4 subcategories
6 SINs in 3 subcategories
Source: GSA's MAS Available Offerings Refresh 34 workbook, on the SAM.gov notice for 47QSMD20R0001 posted November 12, 2026.
Shaded by obligations on those contracts' orders. Select a state for its top contractors.
States by obligations
13,187 listed contracts. The state is the contractor's address on its GSA eLibrary listing, not where the work is performed. Puerto Rico appears in the list but not on the map. 51 list an address the map does not cover, such as a Canadian province. 36 list no state. Obligations are from USAspending, summed per contract across all its orders.
After the consolidation, these are the schedule clauses GSA lists as remaining, the five new ones among them.
01
When an order buys an LLM, generative assistant, chatbot, agentic system or LLM-enabled tool where LLM functionality is a material feature and Government Data goes to or comes from the LLM. It is self-deleting when LLM use stays in the contractor's own back-office tools, or is incidental to a product whose purpose is not AI.
02
The Government owns its Data Inputs, Data Outputs and Custom Developments. The contractor may use them only to perform, for de-identified security telemetry, for support and maintenance, and as the contracting officer authorizes in writing. Training or fine-tuning a model on them, using them for marketing or business decisions, and selling them are prohibited.
03
Restrict access to authorized personnel, log any human access, keep Government Data logically segregated, keep it on the agreed premises or FedRAMP-authorized services, and at the end of the order delete it along with fine-tuned weights, embeddings, indexes and caches.
04
Maximize use of LLMs developed, managed and operated in the United States that no foreign government can control or compel to disclose Government Data; none may be run by an entity an adversary foreign government controls. Disclose within 30 days of award, or of learning of it, any modification made at a non-U.S. government's request or to comply with non-U.S. law.
05
Where an LLM uses reasoning, retrieval or agentic steps, make a summary of the recorded steps available, including retrieval sources and any human-reviewer access. A summary of recorded steps is enough; a faithful account of internal reasoning is not required.
06
Support export of Government Data, interaction history and knowledge bases in open or standard machine-readable formats, with documented APIs, and no format, cost or license barrier to migrating to another service.
07
Flow the clause to subcontractors doing AI design, development, deployment or operation and monitoring under the NIST AI RMF. Compliance can be shown with flowdowns, attestations and existing artifacts such as FedRAMP packages, SOC 2 reports, ISO/IEC 27001 certifications and model or system cards.
08
The clause supplements FAR 52.212-4 and GSAR 552.212-4, and it prevails over conflicting commercial terms, license agreements, terms of service and click-wrap.
Notify the contracting officer after actual knowledge of a material violation of the clause.
Notify the contracting officer and security contacts of an incident affecting Government Data, then send daily updates until resolved, and file CISA's incident form.
Keep relevant logs, forensic images and incident artifacts after a security incident involving Government Data.
When a third-party provider retires a model you do not control, notify the contracting officer and make best efforts to provide access to its successor.
Report a change that materially increases output bias, weakens guardrails or degrades performance or truthfulness.
Give written notice before adding or changing an LLM or its developer, operator, integrator or provider, or a change to the services handling Government Data.
Before replacing a model, use reasonable efforts to give concurrent access to its successor for evaluation: 30 days for a major version, 15 for a minor one.
Disclose the LLMs that process Government Data, by the date the order sets or within 120 days of starting work.
MAS labor categories for large language model, generative, agentic, prompt engineering and chatbot work, the work nearest the clause's trigger, each kept only where the contractor's own price list describes AI work. 11 contracts carry 20 such roles, every one a small business, in GSA CALC+ as of November 15, 2026. The description under each role is the contractor's, quoted from its GSA price list. A labor category is a role the contractor can sell at that ceiling rate, not a role an agency has bought.
47QTCA22D006F · current period to March 31, 2027
Bachelors · 6 yrs · 54151S
Develops and deploys Generative AI models using deep learning frameworks, large language models (LLMs), and NLP pipelines. Fine-tunes foundation models for specific domains, develops APIs for inference, and integrates models into cloud or on-prem environments. Collaborate with data scientists and DevOps teams to deploy scalable, secure AI pipelines. Participates in model evaluation and documentation of ethical/accuracy considerations. Supports prompt engineering and experimentation workflows in support of production readiness.
Bachelors · 8 yrs · 54151S
Designs end-to-end Generative AI systems for real-world applications, leading to the development of reusable model pipelines, embeddings, and retrieval-augmented generation (RAG) systems. Guides teams on selection and adaptation of transformer models, oversees data preparation and prompt engineering strategies, and ensures models comply with privacy and fairness constraints. Works across cloud and hybrid environments to integrate AI services into mission-critical systems. Collaborates with security, UX, and compliance leads to responsible AI implementation.
Bachelors · 11 yrs · 54151S
Serves as technical lead for enterprise GenAI initiatives. Oversee architecture of multi-modal models, including text, image, and code generation. Establishes scalable model governance, monitoring, and retraining strategies. Defines standard practices for evaluation metrics, hallucination reduction, and bias mitigation. Partners with product, legal, and infrastructure teams to operationalize GenAI across departments or programs. Contributes to an internal knowledge base and mentors GenAI developers across multiple teams.
Bachelors · 14 yrs · 54151S
Acts as the principal technical strategist for GenAI across the enterprise. Defines long-term roadmaps for LLM integration, tooling, and infrastructure investment. Evaluates open-source and proprietary model ecosystems, leads cross-functional advisory groups, and influences AI policy and risk governance at the executive level. Authors reference architecture, white papers, and innovation strategies. Provides directions to technical and executive teams on AI capabilities, limitations, and operational impact on scale.
No orders reported on the contract in USAspending yet.
47QRAA26D005Y · current period to May 6, 2031
Bachelors · 3 yrs · 54151S
Develops and integrates artificial intelligence, machine learning, and data-science capabilities for analytical and operational applications. May perform supervised/unsupervised learning, NLP, large-language-model integration, retrieval-augmented generation, agentic workflows, computer vision, anomaly detection, predictive analytics, statistical modeling, feature engineering, data preparation, model evaluation, inference integration, prompt/knowledge engineering, and MLOps.
Bachelors · 4 yrs · 54151S
Designs, develops, integrates, evaluates, deploys, and maintains generative-AI and agentic-AI applications and workflows. Develops LLM-powered applications, AI agents, retrieval-augmented generation (RAG) solutions, tool/function-calling workflows, multi-agent systems, autonomous and human-in-the-loop workflows, memory and context-management mechanisms, and integrations with enterprise applications, APIs, databases, and external tools. Implements agent orchestration, planning, task decomposition, prompt and knowledge engineering, guardrails, security controls, evaluation frameworks, observability, monitoring, and performance optimization. May use technologies and protocols such as MCP, A2A, LangChain, LangGraph, CrewAI, or equivalent frameworks.
Bachelors · 4 yrs · 54151S
Designs, builds, deploys, and operates platforms and pipelines that productionize AI/ML, generative-AI, LLM, and agentic-AI capabilities. Implements model, agent, and data pipelines; model/agent registries; prompt and agent versioning; CI/CD/CT; containerization; orchestration; inference and agent services; GPU/cloud infrastructure; evaluation datasets; tracing; observability; monitoring; drift and performance tracking; access controls; guardrails; and reproducible environments. Integrates models and agents into applications and supports secure, scalable lifecycle management for models, prompts, RAG pipelines, agents, tools, and multi-agent applications from experimentation through production.
Bachelors · 8 yrs · 54151S
Leads the architecture, development, integration, evaluation, deployment, and operationalization of complex generative-AI and agentic-AI solutions. Defines agent and system architectures, model and tool selection, orchestration patterns, RAG and memory strategies, multi-agent coordination, human-oversight workflows, evaluation and red-team approaches, security and guardrails, observability, deployment, and scaling. Integrates agents with enterprise systems, cloud platforms, APIs, databases, and mission data sources; establishes engineering, reliability, and governance practices; leads technical reviews; resolves complex production issues; and provides technical direction to multidisciplinary AI, software, data, and platform teams.
No orders reported on the contract in USAspending yet.
GS35F178CA · current period to February 8, 2030
Bachelors · 1 yrs · 54151S
Provides entry-level support in developing software and generative AI applications. Assists with coding, testing, and integrating features and AI models under supervision. Supports debugging, documentation, and deployment tasks within development environments.
Bachelors · 3 yrs · 54151S
Supports development of software and generative AI applications with limited supervision. Writes and tests code, integrates AI models and APIs, and assists with debugging, optimization, and deployment. Contributes to development standards and continuous improvement efforts.
Bachelors · 5 yrs · 54151S
Designs and develops software and generative AI applications. Builds and integrates features, models, and APIs, optimizes performance, and ensures code quality and security. Leads development efforts and provides guidance to junior developers.
5 orders on the whole contract, $1.2M obligated (USAspending, not only these roles).
47QTCA24D00CD · current period to July 7, 2029
High School · 1 yrs · 54151S
Develops and utilizes simulations and models. Evaluates tools, technologies, architectures, models, and test results. Heads meetings regarding AI with external parties and internal business-embedded innovation teams. Collaborates with lead engineers to define subsystem requirements and acceptance criteria.
High School · 5 yrs · 54151S
Develops and utilizes simulations and models. Evaluates tools, technologies, architectures, models, and test results. Heads meetings regarding AI with external parties and internal business-embedded innovation teams. Collaborates with lead engineers to define subsystem requirements and acceptance criteria.
No orders reported on the contract in USAspending yet.
47QTCA25D004M · current period to January 29, 2030
Bachelors · 4 yrs · 54151S
Designs, tests, and optimizes prompts and workflows for generative AI and large language model applications. Collaborates with technical and business stakeholders to improve AI output quality, accuracy, and usability.
No orders reported on the contract in USAspending yet.
47QTCA20D0078 · current period to March 12, 2030
Bachelors · 3 yrs · 54151S
Designs, tests, and optimizes prompts and interaction patterns for generative AI tools used in training content development and delivery.
4 orders on the whole contract, $84K obligated (USAspending, not only these roles).
GS00F0017Y · current period to March 4, 2027
Bachelors · 3 yrs · 512110, 541219, 54151S
Role: Designs, develops, deploys, and optimizes large language model solutions supporting mission needs. Responsibilities: Fine-tunes and evaluates models; develops data pipelines; ensures scalability and reliability; documents solutions; supports responsible AI practices; and advises stakeholders on AI implementation.
58 orders on the whole contract, $90.2M obligated (USAspending, not only these roles).
47QTCA25D00FH · current period to September 22, 2030
Bachelors · 4 yrs · 518210C
Designs Cloud, builds, and trains conversational agents, chatbots, and NLP-driven interaction systems. Develops conversation flows, intent models, and integrated multi-channel experiences.
No orders reported on the contract in USAspending yet.
47QTCA26D001N · current period to November 20, 2030
Bachelors · 7 yrs · 54151S
The IT Large Language Model (LLM) Specialist is a cutting-edge AI engineer focused on the development, refinement, and application of generative pre-trained transformers (GPTs) and other large language models that exhibit human-like text understanding and generation capabilities. Their core responsibility involves the creation and refinement of LLMs tailored to perform specific and complex natural language tasks, often through fine-tuning on domain-specific datasets or specialized training regimes. They are adept at employing advanced model optimization techniques such as quantization, pruning, knowledge distillation, and regularization strategies to enhance model accuracy, reduce computational overhead, and improve inference performance. A significant aspect of their role involves continuous research and exploration of novel approaches to language understanding and generation, often contributing to the advancement of the field. They are instrumental in applying LLMs to a diverse array of real-world applications, including sophisticated content generation (e.g., creative writing, code generation), nuanced sentiment analysis, highly accurate machine translation, intelligent conversational agents, and robust text summarization. Close collaboration with cross-functional teams is vital for the seamless integration of LLMs into existing systems, applications, and workflows, often requiring the development of robust APIs and deployment strategies. Staying at the forefront of the rapidly evolving language modeling landscape and incorporating cutting-edge research findings into their work is paramount. Furthermore, meticulous documentation of model development, experimental setups, and performance benchmarks is maintained for version control, reproducibility, and compliance.
No orders reported on the contract in USAspending yet.
47QTCA26D005T · current period to June 8, 2031
Bachelors · 5 yrs · 518210C
Develops and tunes generative AI models deployed on cloud platforms for intelligent automation.
No orders reported on the contract in USAspending yet.
47QTCA26D001T · current period to November 24, 2030
Bachelors · 3 yrs · 54151S
Create the Business Specification Documents after gathering the business requirements and understanding the existing chatbot code. Create different JIRA items for the implementation of the application using Agile Software Development methodology. Deploy applications in various environments such as Dev (Development), UAT (User Acceptance Testing), SIT (System Integration Testing) and PROD (Production). Protecting sensitive information by implementing code that mask Personally Identifiable Information (PII).
No orders reported on the contract in USAspending yet.
At least 166 MAS contracts (137 small) carry a labor category whose title names an AI or machine learning role. These are AI roles in general, not the clause's scope; an order still brings the clause in only when it buys an LLM system.
contracts · MLOps and AI platform
4 titles · median $170.33/hr
contracts · AI governance, ethics and risk
31 titles · median $187.17/hr
contracts · AI data science
30 titles · median $214.10/hr
contracts · Machine learning engineering
140 titles · median $173.97/hr
contracts · AI engineering and development
227 titles · median $190.84/hr
Source: GSA CALC+ ceiling rates and each contractor's GSA price list, refreshed weekly; titles and descriptions as each price list states them. CALC+ answers at most 10,000 results a search, so these counts are a floor.
GSA's description of these SINs names artificial intelligence, machine learning or chatbots. The clause still turns on what an order buys, so an LLM offering under any SIN can bring it in.
The clause leans on federal frameworks a seller should have in hand before an order lands.
SIN 561210FAC
contracts GSA eLibrary lists under it
SIN 54151HACS
contracts GSA eLibrary lists under it
The 277 SINs from GSA's offerings workbook, each with its category, NAICS codes, PSC, maximum order, GSA's own description and Resources links, and how many contracts GSA eLibrary lists under it.
Every schedule contract eLibrary lists, one row each, with its other SINs and socio-economic status, filtered to small business or to one SIN, and sorted by obligations, task orders, last date to order or most SINs.
Open a contract to see its task orders, BPAs and BPA calls from USAspending, with obligations, and its price list as GSA eLibrary publishes it, cross-checked against the ceiling rates in GSA CALC+.
The MAS notices and attachments on SAM.gov, so a refresh's category attachments and SF30 can be opened from the vehicle page.
Read from
What the schedule is, what Refresh #34 changes, when the AI clause applies and where GovConDash reads each figure.
MAS is GSA's schedule contract program, run by the Federal Acquisition Service under one solicitation, 47QSMD20R0001. It is organized into 12 large categories and 277 Special Item Numbers that agencies order from.
The update to the MAS solicitation GSA posted on SAM.gov on November 12, 2026. Its SF30 attachment lists the GSAR Part 538 overhaul, the new GSAR 552.239-7001 AI clause, the optional domestic end product certificate at 552.238-130, the nitrile glove change and updates to SINs 561210FAC and 54151HACS. Existing contractors accept the mass modification that applies it no later than 90 days after it is issued, and its changes apply to orders issued after the modification's effective date.
No. By its own terms it applies when an order buys an LLM, chatbot, agentic system or LLM-enabled tool where LLM functionality is a material feature and Government Data is submitted to or produced by the LLM. It is self-deleting when a contractor uses an LLM only in its own internal tools, or when LLM functionality is incidental to a product whose purpose is not AI, unless the contracting officer says otherwise.
In GSA's RGO-2026-01 class deviation on acquisition.gov, which carries the Revolutionary GSAR Overhaul text. The prescription is at 539.71 and the clause at 552.239-7001, dated September 2026 as a GSAR deviation. Formal rulemaking with notice and comment follows the deviations.
The SIN list comes from the MAS Available Offerings workbook GSA posts with each refresh. Today it is Refresh 34, from the SAM.gov notice posted November 12, 2026, and it is rebuilt when GSA posts the next one. Contractor lists come from GSA eLibrary and orders from USAspending, refreshed weekly. Each contractor's price list is read from GSA eLibrary and cross-checked against its ceiling rates in GSA CALC+.
Anyone can open the GSA MAS vehicle page in the Contract Vehicles tab, with its SINs, contractors and orders. Pro and Max add the rest of the platform: your own portfolio and Contracts Report, opportunities matched to your codes, teaming and agency research, and the AI research agents.