This week’s national math graduation exam results show improvement over last year, particularly on the general-level (B) exam. Still, better results do not erase the systemic challenges facing math education in Lithuania – challenges reflected in both national exam data and international research. A new generation of EdTech developers is now working alongside teachers to tackle them.
Results Are Improving, but Challenges Remain
According to Lithuania’s National Agency for Education (NŠA), 85.1% of the 14,011 students who took this year’s advanced-level (A) national math graduation exam passed. Of all test takers, 210 graduating students, or 1.5%, earned a perfect score of 100. On the general-level (B) exam, 77.1% of the 8,009 students who took the test passed. Beginning this school year, all national graduation exams have a uniform passing threshold of 25 points out of 100.
The B-level results improved significantly compared with last year, when just 57.1% of 5,954 test takers passed. Yet nearly one in four graduating students who took the B-level math exam still failed this year – an important data point when assessing the broader challenges facing math education.
Looking at a longer time horizon, math exam results have fluctuated considerably over the past five years. In 2022, 35% of students failed the national math graduation exam – the worst result in the exam’s history. In 2023, 84.6% passed, rising to 89.5% in 2024. In 2025, the national math graduation exam was split into separate advanced (A) and general (B) exams, making direct year-over-year comparisons methodologically difficult. Even so, the broader trend is clear: a growing share of graduating students are passing the exam.
Beyond Exam Results: A Systemic Divide
Education experts emphasize that math is cumulative. A learning gap left unaddressed in an earlier grade creates further gaps as students advance, with the effects compounding all the way through the graduation exam. More than perhaps any other subject, math performance reflects not only a student’s individual effort but also the learning environment around them.
The 2022 PISA results make this divide particularly clear. Fifteen-year-olds in urban areas scored 71 points higher in math than their peers in rural areas – a difference experts estimate is equivalent to nearly two years of schooling and far above the EU average gap of 46 points. Students from socioeconomically disadvantaged backgrounds are also more than four times as likely to fall below the minimum proficiency level in math: 46.5%, compared with 11% of their more affluent peers.
Mathematical Thinking Is a Strategic Investment in Lithuania’s Future
These figures are not simply a domestic education issue. Countries around the world are competing to develop the next generation of talent, and the rapid rise of artificial intelligence (AI) is accelerating that competition. International assessments such as PISA offer one indication of how well countries are preparing young people for the challenges ahead.
Mathematical thinking – the ability to formulate a problem and test the logic of a solution – is becoming one of the most critical skills for the future. Investing in more effective math education is therefore not only an educational priority, but a strategic one.
AI in Math Education: Lessons from 1,000 Teachers and 13,000 Students
To help address this challenge, the philanthropic Devbridge Foundation launched LEARN on.AI. During the 2025–2026 school year, the initiative engaged more than 1,000 teachers and 13,000 students from schools in 44 cities and regions across Lithuania. The program gave teachers and students hands-on experience using AI in education, with a particular focus on schools outside Lithuania’s major urban centers.
“AI is becoming increasingly important across nearly every area of life – from business and science to the public sector. Failing to engage with this shift in education would be a mistake. Every day, we see AI transforming workflows and decision-making, and we invest in businesses built around AI,” says Viktoras Gurgždys, co-founder of the philanthropic Devbridge Foundation.
Viktoras Gurgždys, photo by Agnė Bekeraitytė-Popierė
“Of course, AI also raises a broader question: How will people learn in the future, and what knowledge will matter most? The answers will become clearer over the next few years. But students are learning today. Our goal is to make that learning more effective and help teachers and students adopt new tools together – not as answer machines, but as tools for thinking.”
The Expert View: Where Is Math Education Falling Short?
What does a “tool for thinking” look like in practice? The initiative’s partners, which develop AI-powered math tools, point to specific challenges their solutions are designed to address.
“In a typical classroom, it is physically impossible for a teacher to closely track every student’s individual progress at the same time. This is where well-designed AI can become a valuable assistant. It does not replace the teacher – and it should not. But it can save teachers time, identify learning gaps earlier, and use data to provide a clearer picture of where students are succeeding and where they need additional support,” says Emilija Platūkytė, co-founder of Alfa Erdvė.
Emilija Platūkytė
Mantas Vičius, co-founder and CEO of Elicėjus, says another systemic challenge in Lithuania is that teachers often discover gaps in a student’s knowledge only after seeing the results of a test or exam.
“The challenge is even more acute outside major cities, where schools often face shortages of both subject teachers and student support specialists. That is why assessment and diagnostics should not be a one-time source of stress before an exam. They should serve as an everyday compass for learning – helping teachers and students identify gaps early and focus their efforts where they are needed most,” Vičius says.
Using AI Tools Effectively in Education
According to the 2024 Teaching and Learning International Survey (TALIS), 39% of Lithuanian teachers have already used AI in teaching, compared with 36% across OECD countries. Teachers see the greatest value in adapting learning materials to different student needs (54%) and providing individualized support (44%). At the same time, 76% of teachers who have not yet used AI cite a lack of knowledge or skills as the primary barrier.
A separate 2026 study commissioned by Devbridge Foundation and conducted by Kaunas University of Technology – the first in Lithuania to examine teachers’ real-world experiences with specific AI tools in the classroom – suggests that educators are most likely to adopt new technologies when they reduce workload without adding complexity.
“These figures point to one clear conclusion: teachers do not value AI simply because it is AI. They value tools that help solve specific pedagogical challenges, such as adapting content and personalizing support,” says Vičius.
Mantas Vičius
“General-purpose AI can provide an answer, but it does not understand Lithuania’s math curriculum or know exactly where students tend to struggle within a specific topic. Specialized educational tools are developed alongside teachers and built on evidence-based methodologies. For example, Elicėjus uses the same Item Response Theory methodology applied in OECD PISA assessments. The system can pinpoint a student’s specific learning gaps with a high degree of precision.”
During the 2026–2027 school year, LEARN on.AI plans to reach more than 20,000 students at approximately 70 schools across Lithuania, with a particular focus on schools outside major cities. Schools can choose the solution that best fits their needs – Alfa Erdvė or Elicėjus. Each provider independently manages school selection and license allocation, prioritizing schools outside major urban centers.
Registration: https://devbridgefoundation.org/lt/initiatives/learn-on-ai/