Photomath and Symbolab share the lead, frequently recommended for step-by-step math problem solving and worksheet scanning. When educators search for mastery learning with strong teacher controls, Khanmigo is the usual answer.
Since July, pyBKTrising has become the usual answer for knowledge tracing inquiries, while EduKTMrising and Symbolabrising remain steady fixtures across specialized prompts.
Latest insights
Weeks of Aug 31 – Sep 27, 2026 · Since Jul 5
Microsoft Math Solver Gaining: Microsoft Math Solver now leads Wolfram|Alpha in AI Tutoring and Study Platforms, #5 to #6
Microsoft Math Solver Gaining: Microsoft Math Solver ranked #6 in AI Tutoring and Study Platforms last week and #5 this week
Mathway Gaining: Mathway ranked #8 in AI Tutoring and Study Platforms last week and #7 this week
Wolfram|Alpha Slipping: Wolfram|Alpha dropped out of the AI Tutoring and Study Platforms top ten
Wolfram|Alpha Slipping: AI answers in AI Tutoring and Study Platforms name Wolfram|Alpha less often: mention rate declined from 28.1% to 21.4%
30% of citations point at the ranked brands’ own sites. arxiv.org, github.com, youtube.com are the most-cited sites outside the ranked brands’ own domains.
Yes. There are several organizations worth looking at, but an important distinction is that few companies sell “knowledge tracing” as a standalone API. More commonly, KT is embedded in an adaptive-learning platform.
Commercial options
Provider
KT / personalization approach
Best fit
Knewton
Learner knowledge/proficiency modeling + knowledge graph + adaptive recommendations. Its Enterprise platform exposes APIs for integrating recommendations into your own learning product.
Best starting point if you want to build your own product around an external adaptive engine
I need to find tools that offer knowledge tracing capabilities to support personalized student learning journeys.
What is the best math solver app that shows the steps for solving a problem?
Which AI tutor supports mastery learning with strong teacher controls?
Riiid
Socra
Deep-learning-based learner modeling, prediction and personalized curriculum. Riiid's technology has been used to analyze learner weaknesses and recommend customized questions/courses.
Test prep, large-scale AI tutoring, B2B partnerships
ALEKS
Knowledge Space Theory, rather than conventional BKT/DKT. It maintains a detailed learner knowledge state and selects topics the learner is ready to learn.
Math/STEM, mastery learning
Carnegie Learning
MATHia is an adaptive mathematics platform that continuously models student performance and adjusts instruction.
K–12 math
Squirrel AI
AI-driven adaptive learning with fine-grained modeling of learner mastery and personalized paths.
K–12 / large-scale adaptive education
CENTURY Tech
AI-driven learner modeling and personalized learning paths.
Schools / institutional deployments
If you specifically want the algorithm, rather than a finished learning platform
Two open-source projects are particularly useful:
pyBKT — Python implementation of Bayesian Knowledge Tracing, including variants with individualized priors, learning rates, item-level guessing/slipping, etc. It can be trained directly on your own student interaction data.
pyKT — PyTorch toolkit containing 10+ deep knowledge-tracing models, including DKT, DKT+, DKVMN, SAKT, SAINT, AKT, HawkesKT, etc. It's better if you're evaluating modern neural KT models rather than just classical BKT.
What I'd consider for your use case
If you're building an AI personalized-education product, I'd divide the market into three approaches:
1. Buy the adaptive engine → Knewton
You send learner interactions/content information and receive personalized recommendations. This is closest to “I want KT capabilities without developing the whole thing myself.” Knewton explicitly documents an API-based integration model.
2. Build the KT layer yourself → pyBKT / pyKT
This gives you much more control over your definition of mastery, curriculum graph, cold-start behavior, forgetting, explainability, and recommendation logic. pyBKT is particularly attractive for an initial production prototype because BKT is relatively interpretable.
3. Buy a complete adaptive-learning ecosystem → ALEKS / Carnegie Learning / others
These are more appropriate if you're buying an education product, rather than technology to embed into your own product. ALEKS, for example, has a sophisticated proprietary knowledge-state system but isn't simply selling you its underlying KT algorithm.
One subtle but important point: Knowledge Tracing isn't itself the personalized-learning path algorithm. KT estimates “what does this learner know right now?”; a separate sequencing/recommendation policy decides “what should they learn next?” Knewton, for example, explicitly combines learner knowledge-state estimates with recommendation logic.
If you're evaluating vendors for a new AI personalized-education platform, I can also give you a 2026 vendor comparison of ~10 providers, including API availability, BKT vs DKT, knowledge graphs, pricing/business model, integration options, and whether you can bring your own curriculum/data.
If you are building personalized education paths, Knowledge Tracing (KT) algorithms serve as the mathematical backbone for tracking how a student's mastery of skills evolves over time.
Several prominent open-source libraries, academic toolkits, and platforms provide ready-to-use Knowledge Tracing implementations ranging from classical probabilistic methods to modern deep learning architectures.
1. Dedicated Open-Source Python Libraries
If you want to implement and train models yourself using Python, these two toolkits are the industry standard for research and deployment:
pyKT (Deep Learning Knowledge Tracing) : Built on top of PyTorch, pyKT is a comprehensive library designed to benchmark and train deep learning-based KT models. It comes with standardized preprocessing for major education datasets and implements over 10 popular neural architectures (such as DKT, DKVMN, and AKT).[](https://github.com/pykt-team/pykt-toolkit)[[2](https://github.com/CAHLR/pyBKT)](https://pykt.org/)
If you prefer an enterprise platform or API rather than coding models from scratch, several educational technology infrastructure providers embed knowledge tracing and student modeling into their suites:
Riiid : An AI solutions provider focused heavily on education, offering proprietary AI tutoring and deep learning knowledge-tracing models optimized for test prep and enterprise learning.
Knewton (by Wiley) : One of the pioneers in adaptive learning platforms, offering foundational infrastructure and adaptive engines that trace student proficiency dynamically.
Open edX : While primarily a Learning Management System (LMS), it supports extensions and XBlocks where custom adaptive learning algorithms and student progress-tracking metrics can be integrated.
Are you looking to build a custom model from scratch using Python libraries like pyBKT/pyKT, or are you searching for an enterprise API/platform to integrate directly into an existing product?
For personalized learning paths, knowledge tracing (KT) algorithms estimate what a learner knows, how their mastery changes over time, and which concept or activity they should tackle next. If you're evaluating software providers, it's useful to distinguish between vendors that offer KT as a built-in feature, platforms that provide adaptive learning engines, and tools that let you implement the algorithms yourself.
The main algorithms to look for include:
Bayesian Knowledge Tracing (BKT): Models the probability that a learner has mastered a skill based on their responses.
Deep Knowledge Tracing (DKT): Uses recurrent neural networks to learn patterns from sequences of learner interactions.
Transformer-based knowledge tracing: Uses attention mechanisms to model longer interaction histories.
Item Response Theory (IRT) and performance models: Estimate learner ability and question difficulty; related to KT, but not equivalent to tracking mastery over time.
Performance-factor and logistic models: Predict future success using factors such as prior attempts, correctness, and practice history.
Providers and tools to evaluate
I'd divide the options into two groups: tools for implementing KT directly in your own product, and commercial learning platforms with built-in adaptive intelligence.
1. Direct algorithm implementations
These are the best starting points if you want control over your learner model, recommendations, and learning-path logic.
pyBKT
(Best for interpretable mastery estimates)
An open-source implementation of Bayesian Knowledge Tracing. A good option when you want to track mastery at the individual skill level and explain how evidence updates the learner's estimated knowledge.
Website:
pyKT
(Best for comparing deep-learning models)
A PyTorch toolkit for training and benchmarking many knowledge-tracing models, including DKT, attention-based models, and transformer-style approaches. Particularly useful for experimentation and model selection.
An implementation-oriented option for experimenting with KT models. Related repositories in this ecosystem include tools for processing educational datasets, which can help you prepare learner interaction histories for model training.
Website:
These are primarily developer libraries rather than managed commercial services. You generally provide your own learner event data, train or configure the models, and deploy the inference pipeline.
2. Commercial adaptive-learning providers
These companies are worth evaluating if you want an existing learner-modeling and personalization system rather than implementing everything yourself.
Knewton
Offers adaptive-learning technology designed to personalize educational content using learner proficiency and content relationships. Investigate its current enterprise integration options if you want to embed adaptive recommendations in your own application.
Website:
Riiid
Develops AI-driven education technology for learner-performance prediction and personalized study recommendations. A potential fit for data-driven adaptive learning, although you should confirm whether its current offerings support licensing or embedding the underlying models.
Website:
ALEKS
Uses adaptive assessment and knowledge-state estimation to identify gaps and guide learners toward topics they need to study. It is more relevant as a reference platform or integrated educational solution than as a confirmed standalone KT API.
Website:
CENTURY Tech
Provides AI-powered personalized learning that adapts to learner needs. Consider it if your goal is to use an established adaptive-learning environment; verify the extent of access to its underlying learner models and APIs.
Website:
Important distinction: Commercial platforms may advertise adaptive learning without exposing a specific BKT, DKT, or transformer-based KT model. The named vendors above are candidates for evaluation, not a claim that every vendor licenses a standalone knowledge-tracing algorithm.
Which should you choose for your learning-path engine?
Your priority
Recommended starting point
Simple, interpretable skill mastery
pyBKT
Compare DKT, attention, and transformer models
pyKT
Build your own model using educational datasets
pyKT + EduData
Integrate an existing commercial adaptive-learning engine
Evaluate Knewton and Riiid
For a product you are building yourself, I'd start with pyBKT as a baseline, then benchmark pyKT models against your actual learner data. This lets you establish whether the complexity of deep learning produces enough improvement to justify the added operational cost.
How to connect knowledge tracing to personalized paths
Choose a new concept, targeted practice, review, or assessment
The key architectural point is that knowledge tracing and learning-path selection are separate problems. KT estimates the learner's knowledge state; your sequencing algorithm uses that state to decide what to teach next.
Before selecting a provider, confirm whether it supports real-time inference, per-skill mastery estimates, your own content taxonomy, API integration, and export of learner data. These capabilities will matter more than the vendor's use of the term “AI.”
One question that would help me narrow the list: Are you building your own software and want an embeddable algorithm/API, or are you looking to license a complete adaptive-learning platform?
Commercial platforms offer engine integrations, while specialized python libraries provide the actual algorithms (like Bayesian or Deep Knowledge Tracing).
1. Adaptive Learning Engines (B2B / API-Driven)
If you prefer an out-of-the-box engine that handles student modeling, content sequencing, and tracking via standard protocols (like xAPI), these providers license their engines for integration:
Core Tech: Provides an adaptive learning engine that integrates into existing LMS/content platforms via xAPI standards. It tracks student ability, manages user profiles, and sequences diagnostic questions dynamically.[](https://www.adaptemy.com/engine/)
pyKT: A robust, PyTorch-based deep learning library explicitly designed for knowledge tracing. It standardizes state-of-the-art DKT (Deep Knowledge Tracing) models, data splitting, and parameter optimization.[](https://pykt.org/)
EduKTM: An open-source knowledge tracing toolkit designed to bridge educational psychology with machine learning, supporting traditional and deep models (DKT, AKT, etc.).
3. Architectural Approach: The DIY Stack
If you choose to build your own using standard backend infrastructure rather than a black-box engine, the typical stack involves:
Data Capture Layer: Tracking granular interaction events (e.g., correct/incorrect, response time, hints requested) using xAPI (Experience API) specs.
