Subscribe
Reports

AI Tutoring Benchmarks

What each one measures, who it’s for, and how to use it, updated as the field evolves

November 16, 2026
AI tutoring benchmarks

As AI tutors get better at explaining, the harder question is whether they actually help students learn. A growing number of researchers, companies and nonprofits are building benchmarks to answer that, from tests of whether a tutor gives hints instead of answers to studies that measure learning gains with real students.

This page tracks the open benchmarks we find most useful for founders building AI tutors and learning tools: what each one measures, the grade levels and subjects it covers, and how you can access it. We’ll keep it updated as new ones appear.

For founders building AI tutors: check out our take on what these benchmarks capture, what they miss, and the five questions you should be ready to answer in this article: AI Tutors Can Teach. Who Will Learn?

Know of a benchmark we missed, or building one yourself? Reach out to tony[at]reachcapital[dot]com.

Last updated: November 2026

BenchmarkCreatorWhat it measuresLevelPublished as
TutorBenchScale AITeaching moves: explaining, feedback, hints without giving away answersHS / AP STEMLeaderboard
MathTutorBenchETH Zurich and partnersCorrectness and teaching movesK-12 mathCode, data
MRBench + BEA 2025MBZUAI and partnersSpotting and fixing student mistakesK-12 mathData
DrawEduMathAllen Institute for AI, Worcester Polytechnic Institute and partnersUnderstanding students’ handwritten and hand-drawn math workK-12 mathCode, data
GradeBenchPensiveGrading handwritten student work against instructors’ rubricsUniversity STEMResults chart
MMTutorBenchAcademicTeaching moves with images and diagramsK-12 mathPaper
TutorMomentsAllen Institute for AI and partnersWhen to scaffold, push for rigor or step back, using real tutoring sessionsGrades 2–7 mathCode, data (preview)
CSTutorBenchUIUCTeaching quality of small models in block-based programmingGrades 6–8 codingRubric, code
Pedagogy BenchmarkFab IncKnowledge of teaching and learning scienceTeacher levelGated data
EduBenchBeijing Institute of TechnologyTeaching knowledge and movesMixedData
ELBenchAcademicTeaching knowledge and soundness, across nine modelsMixedPaper
SafeTutorsAcademicPedagogical soundness: over-disclosure, reinforcing misconceptions, dropping scaffolding over long conversationsK-12 STEMPaper, data
EduZoneAcademicResistance to misuse, including disguised cheating requestsK-12Paper
Safe-Child-LLMAcademicAge-appropriate responsesAges 7–17Code, data
TutorCaseBenchAcademicSupporting students with learning and behavioral challengesSpecial populationsPaper (under review)
EduClaw-BenchAcademicSustained tutoring over 30 days with simulated learnersSimulated learnersPaper
StudentBenchHandshakeLearning gains with real learners, compared with human tutorsAdults (GRE prep)Code, data
Learning Outcomes Measurement SuiteOpenAI, University of Tartu, Stanford SCALELong-term learning outcomes from real-world AI useK-12 and beyondForthcoming