MLJAR AutoML

MLJAR AutoML is an open-source tool for automated supervised learning on tabular data, including classification and regression. It is best suited to analysts and Python users who want to automate parts of a modeling workflow while retaining control over training settings. Its appeal is configurable workflows and reusable experiment reports; look elsewhere for forecasting or time-series modeling.

Overview

Use MLJAR Studio notebooks or the Python package in scripts and notebooks. Depending on mode and configuration, it can preprocess data, select and tune models, engineer and select features, and build ensembles. This can reduce routine setup, but users still need to make informed choices about validation, metrics, algorithms, and time budgets.

Experiments can include model comparisons, validation results, training times, explanations, and saved artifacts, with summaries in Markdown and JSON. Reports and explanations vary with the task, mode, algorithms, and explanation settings, so the output is not uniform across every run.

Key features

Four training modes

Explain, Perform, Compete, and Optuna provide different ways to configure training. Users can set algorithms, validation, metrics, time budgets, and explanations. That flexibility suits projects with specific evaluation needs, though it puts responsibility for those choices on the user rather than hiding them behind a single fixed workflow.

Fairness and deployment

Fairness-aware training evaluates models against a chosen metric and threshold and can try sample reweighting to mitigate disparities. It is a useful check, not a guarantee of a fair result. The Python package also integrates with Mercury to generate apps for single predictions or batch CSV predictions. Apps can be previewed locally, published through MLJAR Platform, or self-hosted with Docker, giving teams deployment options beyond the modeling notebook.

Privacy and support

MLJAR says it uses SSL/TLS encryption, access controls, and authentication mechanisms. It advises against submitting sensitive or regulated data to AI features unless authorized, a relevant caution for projects with strict data-handling requirements. The team invites questions through its contact channel.

Pricing

The mljar-supervised Python package is free under an MIT license. It is the clearest fit for users who want the open-source package for scripts and notebooks; the package plan does not include the separate Mercury app quotas described by the other plans.

Mercury's Free plan is free and includes 50 prompts per month, 10 published conversations, and 1 public Mercury web app. It is suited to small-scale evaluation, but its monthly prompt allowance and single public app constrain ongoing or broader use. Pro costs $20 a month for 500 prompts per month, 50 published conversations, 3 public apps, and 1 private app. Business costs $60 a month for 2,000 prompts per month, 200 published conversations, 10 public apps, and 3 private apps; it fits teams needing more capacity and private app slots.

The $199, paid once Perpetual license is a one-payment option that remains yours forever and includes 1 year of updates, local LLMs with Ollama, and the ability to bring your own provider keys. These plan quotas and app allowances concern Mercury; they should not be confused with the open-source package's modeling capabilities.

Platforms

MLJAR AutoML supports Linux, macOS, Windows, web, and self-hosted use. API access is supported, and deployment mode is batch. Its task scope is supervised tabular modeling: it does not support forecasting or time-series modeling.

Who it's for

Choose MLJAR AutoML if you work with tabular classification or regression and want automated preprocessing, model selection, tuning, and reporting without surrendering control of key settings. It is less suitable when your project depends on forecasting, time-series methods, or a guaranteed fairness outcome.

Pros and cons

• Pros: Four modes and configurable algorithms, validation, metrics, time budgets, and explanations let users adapt training to a project.
• Pros: Markdown and JSON summaries, comparisons, validation results, and saved artifacts support review and reuse of experiment outputs.
• Pros: Mercury can turn models into single-prediction or batch CSV apps, with local preview, platform publication, and Docker self-hosting options.
• Cons: It is limited to supervised tabular tasks, so forecasting and time-series projects need another tool.
• Cons: Reports and explanations depend on run settings, while fairness checks do not guarantee a fair model; users must interpret the outputs and choices.

Alternatives

Browse AutoML Software for more automated modeling options, or Predictive Analytics Software for a broader category.

• AutoGluon is another free, open-source Python library; consider it when Apache 2.0 licensing or its Linux, macOS, Windows, and self-hosted platform mix suits your project better.
• KNIME Analytics Platform is free and open source for Linux, macOS, and Windows; choose it if you want that platform rather than MLJAR's Python package and notebook routes.
• Weka offers free stable and development versions; it may suit users who prefer its desktop-oriented platform options and version choices.
• BigML offers a free plan with unlimited tasks and storage, but a 16 MB maximum dataset size per task, 2 parallel tasks, and 1 user; consider it for hosted use within those limits.
• Orange Data Mining is free software under GNU GPL 3.0 or later for Linux, macOS, and Windows; consider it if that licensing and platform combination fits your workflow.
• Prophet is open-source forecasting software with R and Python implementations; choose it when forecasting is the priority, which MLJAR AutoML does not support.
• Aible offers a free plan with 10 basic questions per month, excluding advanced questions, and Aible-hosted data; consider it if that hosted question-based approach fits better.
• ZebraDesigner is an alternative.

Verdict

MLJAR AutoML is a strong fit for Python users and analysts who need configurable supervised tabular workflows, experiment comparisons, and reusable artifacts. Its open-source package is the main reason to choose it; its narrow task scope and mode-dependent reporting are the reasons to look elsewhere when you need forecasting, time-series modeling, or consistent explanation and fairness outcomes.