# MATHia: reviews and analysis > Secondary maths tutor with graduated hints that will not let you finish a unit by asking for them. - Canonical: https://serchai.com/en/reviews/mathia/ - Site: Serchai (https://serchai.com) — AI tools comparator - Language: en - Updated: 2026-08-05 --- ## Verdict MATHia comes out of twenty-five years of cognitive tutor research at Carnegie Mellon and it shows in what it does well: hints get more specific as you ask for them, and the shortcut of asking for all of them is closed off, because mastery stops rising past a point. What weighs against it is everything else: it is sold to schools and districts only, Carnegie Learning publishes no price figure at all, and recent independent evidence is lukewarm, with an averaged effect size of 0.04 and a state review scoring it 30% on supports for learners. **Best for:** US secondary maths departments that want guided practice with hints and per-student mastery data. **Rating:** 3.1/5 ## Pros - Two kinds of hints, the ones that fire on a common mistake and the ones you request, each more detailed than the last - You cannot finish a unit by asking for hints: mastery stops rising - Its own cognitive tutor engine with twenty-five years of research behind it, not somebody else’s language model with a text box ## Cons - No public price: it sells through demos and sales contact, so a family cannot buy it - The most recent independent evidence is lukewarm, with an averaged effect of 0.04 across three studies - The 2025 Texas state review scores its supports for learners at 30% - A US curriculum from sixth grade to high school: outside that syllabus it does not fit ## Key facts - Price: Price on request: no public rate card, sold via demo - Free trial: No - Platforms: Web - Categories: [Learning](https://serchai.com/en/best-ai/learning/) - Official website: https://www.carnegielearning.com/solutions/math/mathia ## What the internet says (agentic sweep) MATHia is one of the few tools in the segment that can show how it avoids doing the homework for the student: its hints get more specific as they are requested and mastery stops rising if they are abused. The flip side is that Carnegie Learning publishes no price figure on its site, that the independent evidence averaged by Evidence for ESSA lands at an effect of 0.04 across three studies and 20,485 students, and that the 2025 Texas state review scores supports for learners in its supplemental materials at 30%. - Sweep date: 2026-08-05 - Derived score: 3.1/5 ### Axes - Results: 3.2/5 - Control: 3.4/5 - Real price: 2.5/5 - Integration: 3/5 - Support: 3.6/5 ### Recurring themes in favor - The student guide describes two kinds of hint, the one that fires on a common mistake and the one you request with growing detail, and states that a unit cannot be completed by asking for them because mastery stops rising (strong theme) - The engine comes from twenty-five years of cognitive tutor research at Carnegie Mellon and combines symbolic AI with machine learning, so it is not a text box over somebody else's language model (strong theme) - It has genuine independent review, a rarity in this segment: EdReports gave it 100% on its first two gateways and 97% on the third in 2019, and Evidence for ESSA classifies it as strong evidence (present theme) ### Recurring themes against - Carnegie Learning publishes no price figure for MATHia on any official page: the only route is requesting a demo or talking to sales, which shuts out any family (strong theme) - The independent figure averaging three studies and 20,485 students is an effect size of 0.04, well below what the large trial the company cites suggests, a trial that in any case evaluated the predecessor Cognitive Tutor between 2007 and 2013 rather than today's product (strong theme) - The 2025 Texas state review of its sixth grade supplemental materials gives it 30% on supports for all learners, 48% on instructional design and 52% on productive struggle, with full curriculum coverage (strong theme) - There is almost no public user conversation to score it with: Capterra holds a single company review and it is not about MATHia, and the teacher reviews on Common Sense Education are going away because the programme shut down in January 2026 (present theme) ### Sweep sources - [Docs] https://www.carnegielearning.com/solutions/math/mathia — Página oficial de producto para los grados 6 a 12, con «Request a Demo» como única vía y sin ninguna cifra de precio - [Docs] https://www.carnegielearning.com/blog/mathia-ai — Explicación oficial del motor: «symbolic and machine learning-based AI» nacido del trabajo de psicólogos cognitivos e informáticos de Carnegie Mellon hace veinticinco años, sin mención de IA generativa - [Review sites] https://www.evidenceforessa.org/program/carnegie-learning-high-school-math-solution/ — Evaluación de la Universidad Johns Hopkins: calificación «strong» con un tamaño de efecto de +0,04 promediado sobre tres estudios y 20.485 alumnos - [Review sites] https://www.edreports.org/reports/overview/carnegie-learning-math-solution-2018 — Informe independiente del 5 de diciembre de 2019: 100 % en enfoque y coherencia, 100 % en rigor y prácticas matemáticas y 97 % en usabilidad - [Review sites] https://im.tea.texas.gov/programs/texas-supplemental-math-solution-grade-6-student-1-year-license — Revisión estatal de Texas de 2025 del material suplementario de sexto: aprobado por la SBOE con cobertura del 100 % del currículo, 30 % en apoyos al alumnado, 48 % en diseño instruccional y 52 % en esfuerzo productivo - [Press] https://eric.ed.gov/?id=EJ1024233 — Artículo revisado por pares sobre el ensayo aleatorizado del Cognitive Tutor Algebra I: sin efectos el primer año y con efectos positivos el segundo > **TLDR:** MATHia is Carnegie Learning's maths tutor for US grades 6 to 12. Its merit is that the hints are designed not to become the answer, and its documentation says so plainly: asking for help over and over will not finish a unit for you. Its drawbacks are about buying and about outcomes, because it sells to schools with pricing on request and the recent independent evidence is thinner than its pedigree suggests. ## What MATHia is and how it works MATHia did not come out of the chatbot wave. It descends from a line of work that started twenty-five years ago, when a group of cognitive psychologists and computer scientists at Carnegie Mellon teamed up with maths teachers from Pittsburgh Public Schools to build what the literature calls a cognitive tutor: a system that models how a student solves a problem, step by step, and detects the exact moment the reasoning goes off course. That matters for understanding what you are buying. The vendor describes its engine as a blend of symbolic AI and machine learning, and its page on the subject makes no mention of generative models at all. The generative product in the house is a different one, LiveHint AI, and it has been in beta for a while. So MATHia is not a text box over somebody else's model with the logo swapped, which is the number one risk in this niche: it is software with a model of the student inside it. Students work through units organised by skill, and the system measures not how many exercises were finished but how much of each objective has been mastered. That distinction changes what the teacher sees. A completion dashboard tells you who did the homework. A mastery dashboard tells you who understood it, which is the question a maths department actually wants answered. ## What it is like day to day The interesting part of MATHia for the segment it occupies is the design of its hints, and its student documentation spells it out with a candour that is rare here. There are two kinds. The first appear on their own when the system recognises a common error, the sort it has seen thousands of times at that point in the syllabus. The second are requested by the student, and each time they insist they get a more specific version than the last. The sentence that defines the product is what happens if someone tries to abuse that ladder: mastery stops rising past a certain percentage even if hints keep coming, so a unit cannot be completed by asking for help. That is the exact opposite of the shortcut offered by tools that solve a photographed exercise, and it answers the question this segment is built around. It is worth saying where that discipline breaks. In its step-by-step tool, if a student gets a step wrong twice, the system completes that step for them. It is a defensible choice so nobody sits stuck forever, but it is an open door, and anyone using it as an escape hatch will move on without learning. On the teacher's side, the vendor offers a live view of the class showing who is working and who has been idle for a while. That is the kind of data you appreciate when there are thirty people in the room and the time goes into spotting who needs you next. The big constraint is fit. The syllabus is American and the years it covers run from roughly age eleven to the end of secondary school. Outside that curriculum the material stops being aligned to anything, and that is a solid reason for a school elsewhere to look at something else. ## Pricing and plans Here comes the awkward part: there is no published price. Carnegie Learning gives no figure for MATHia on any official page, and the only route it offers is requesting a demo or leaving your details for a salesperson. On this site that is declared as pricing on request, because that is the truth, and it is not padded out with an estimate. Per-student yearly figures do circulate in evaluations and public procurement records outside the vendor, but a price that does not appear on the website of the company charging it is not a published price, so it does not get written here. What can be said is how you buy: by student licence per year, inside a school or district negotiation, which is a process of weeks rather than a card payment. The practical consequence is that a family looking for a tutor for their child this afternoon cannot buy MATHia. It is an institutional product, and that filter matters more than any feature. ## Who it is for (and who it is not) MATHia fits a US secondary maths department that wants guided practice with per-student mastery data and that values a system which resists handing over the answer. The independent evidence behind it is real, which almost nothing else in this segment can show: EdReports gave it 100% on its first two evaluation gateways and 97% on the third in December 2019, and Evidence for ESSA, run out of Johns Hopkins, rates it as strong evidence. Read that same evidence all the way through before signing. The effect size Evidence for ESSA averages across three studies and 20,485 students is 0.04, which is small. The large, famous trial usually cited, funded by the Department of Education, found no effects in year one and effects in year two, but it evaluated the predecessor Cognitive Tutor between 2007 and 2013 rather than the product sold today. And the most recent state review of its materials, from Texas in 2025, approves the programme with full curriculum coverage while scoring it 30% on supports for all learners, 48% on instructional design and 52% on productive struggle. It is not for families, not for anyone following a curriculum other than the American one, and not for anyone who needs to decide today. For the first of those, [Khanmigo](https://serchai.com/en/reviews/khanmigo/) covers the home case with a subscription and the same resistance to handing over answers. To try something at no cost, [CK-12 Flexi](https://serchai.com/en/reviews/ck12-flexi/) is free, though it answers with less ceremony. The segment ranking is in [the best AI tutors for learning](https://serchai.com/en/best-ai/ai-tutors-for-learning/) and the wider map is in [the best AI tools for learning](https://serchai.com/en/best-ai/learning/). ## Alternatives to MATHia If what draws you is the Socratic discipline rather than an institutional contract, [Khanmigo](https://serchai.com/en/reviews/khanmigo/) is the direct comparison and takes minutes to set up. If the budget is zero, [CK-12 Flexi](https://serchai.com/en/reviews/ck12-flexi/) gives free tutoring over its own material, with the caveat that it hands over the direct answer by default. And for studying from your own notes rather than a fixed syllabus, that ground belongs to [Microsoft Study and Learn](https://serchai.com/en/reviews/microsoft-study-and-learn/). All three sit in [alternatives to MATHia](https://serchai.com/en/alternatives/mathia/). ## Frequently Asked Questions ### Can a parent buy MATHia for their child? Not through any normal route. Carnegie Learning sells to schools and districts and does not even publish a price: you request a demo and speak to a salesperson. ### Does it give the answer if the student keeps pushing? Generally no: hints get more specific but mastery stops rising if they are requested endlessly, so a unit cannot be finished by asking for help. The exception is its step-by-step tool, where after two failed attempts the system completes the step. ### Does it use generative models like the chatbots? No. Its engine combines symbolic AI and machine learning and predates that wave. Carnegie Learning's generative product is a separate one, LiveHint AI, and it remains in beta. ### Does it work outside the US curriculum? It is not built for that. It covers the American syllabus for grades 6 to 12, so outside that system it loses its main advantage, which is arriving aligned to the standards. ### What does the independent evidence say? That it exists, that it is more than almost anyone in this segment can show, and that it is modest: Evidence for ESSA averages an effect of 0.04 across three studies, and the 2025 Texas review pairs full curriculum coverage with weak scores on supports for learners. ## Alternatives - [Khanmigo](https://serchai.com/en/reviews/khanmigo/) — Khan Academy's Socratic tutor for learners and families in the United States. (4.4/5) - [CK-12 Flexi](https://serchai.com/en/reviews/ck12-flexi/) — A non-profit foundation’s tutor: genuinely free, anchored to its own library and too quick to hand over the answer. (3.7/5)