Open Source AI Model for Tutoring Grant

The Open Source AI Model for Tutoring grant is funded by the Bill and Melinda Gates Foundation in partnership with the K-12 AI Infrastructure Program. This initiative seeks to advance the development of an open source artificial intelligence model specifically designed for K-12 math tutoring in the United States. The program is rooted in the recognition that current general purpose AI systems are not optimized for effective teaching and often fail to replicate the pedagogical strategies of human tutors. The funder aims to catalyze innovation by supporting a single large scale project that will produce a foundational, publicly accessible AI model and associated research outputs. The primary objective of the grant is to create an AI tutoring system that can match or exceed the effectiveness of expert human tutors. The funded project must address well documented limitations in existing AI systems, including their tendency to provide answers rather than encourage productive struggle, difficulty identifying student misconceptions, excessive verbosity, and lack of personalization. The model is expected to incorporate pedagogical frameworks, multimodal inputs such as text, audio, and visual data, and align with instructional practices used in real classroom settings. The resulting system should improve student motivation, engagement, and learning outcomes, particularly in mathematics education. Funding for this opportunity is substantial, with up to 8000000 USD available for a single award. The project period is expected to span 30 to 36 months, beginning around November 2026. Allowable costs include personnel, consulting, computational infrastructure, data acquisition and annotation, travel for project activities, and integration work with education technology partners. Unallowable costs include lobbying, pre award expenses, and proprietary product development. All outputs must be released under permissive open licenses, ensuring that the resulting tools and datasets are broadly accessible and reusable. Eligibility is open to organizations and institutions with demonstrated experience in developing and deploying AI models within K-12 education contexts in the United States. Applicants must show prior work predating May 8, 2026, including peer reviewed publications and contributions to digital public goods such as datasets or open source models. Teams are required to include expertise across multiple domains, including AI engineering, K-12 practice, learning science, and education technology partnerships. Projects must also demonstrate prior deployment or evaluation at meaningful scale using real student data. The application process requires submission through a Qualtrics form, accompanied by supplemental materials such as prior work samples, resumes for key personnel, letters of commitment from partners, and a detailed budget spreadsheet. The proposal must address multiple evaluation criteria including significance, team capabilities, project workplan, and dissemination strategy. Each section has specific character limits, and the total narrative must not exceed 23000 characters. Applicants must also disclose any use of AI tools in preparing their submission. Proposals will be evaluated holistically based on their potential impact, technical feasibility, and alignment with program goals. Reviewers will assess the strength of the proposed approach, the qualifications of the team, the robustness of the data acquisition and evaluation plans, and the likelihood that the resulting public goods will be widely adopted. The program emphasizes safety, fairness, and responsible AI practices, particularly given that the model will interact directly with students. Key dates for the opportunity include an RFP release on June 1, 2026, and a submission deadline of July 31, 2026. The grant is not explicitly stated as recurring, and there are no required pre application steps such as a letter of intent for this phase. Applicants can direct questions or request support via the provided email contact. The overall timeline anticipates project kickoff shortly after award notification, followed by phased development, testing, and public release of the AI model and associated resources.

Focus on clearly addressing known AI tutoring limitations with evidence based methods demonstrate strong prior work and include a multidisciplinary team with real world deployment experience