Weka, the machine learning workbench from the University of Waikato in New Zealand, is still free, open source under the GNU GPL, and maintained. Eibe Frank announced Weka 3.8.7 (stable) and 3.9.7 (development) on the mailing list on April 2, 2026, per the release announcement. The main change is clean support for Java 25. Installers now cover Intel and ARM on Windows, macOS and Linux, each with a bundled BellSoft Java 25 runtime that includes JavaFX.
So the reason to move is rarely that Weka stopped working. People outgrow it: they want a friendlier visual canvas, a Python workflow that matches what employers use, bigger data than one Java heap holds, or a platform a whole team can share. The seven tools below cover each of those cases. If you came from RapidMiner, our RapidMiner alternatives page covers that path.
Prices below are as of November 6, 2026. Our method is on How we review tools.
Quick comparison
| Tool | Best for | Price | Platforms |
|---|---|---|---|
| Orange | Visual workflows for teaching and quick analysis | Free, open source | Windows, macOS, Linux (pip/conda) |
| KNIME | Visual pipelines with deep data preparation | Free; Pro €19/mo | Windows, macOS, Linux |
| scikit-learn | Code-first machine learning in Python | Free, open source | Any OS with Python 3.11+ |
| H2O-3 | AutoML and models on data larger than memory | Free, open source | Any OS with Java; Python and R clients |
| Dataiku | A shared platform for analysts and data scientists | Free Edition; paid plans contact sales | Browser; self-install on Mac and Linux |
| IBM SPSS Modeler | A vendor-backed visual modeller with a public price | From USD 529/mo | Windows, Mac |
| RapidMiner AI Studio | Enterprise buyers already on Siemens software | Contact sales | Cloud, on premises or hybrid |
Orange
Weka's Knowledge Flow and Orange share an idea: drop components on a canvas and wire them together. Orange, from the University of Ljubljana, does it with more polish. Scatter plots, trees and confusion matrices update the moment you change a setting, and you can select points in one chart and see them flow into the next widget. For a lecture or a first project, that live feedback teaches faster than Weka's Explorer tabs.
Version 3.40.0 was uploaded to PyPI on December 20, 2025, under GPL v3 or later, and needs Python 3.11 or newer. There are installers for Windows and for Apple Silicon and Intel Macs. See our Orange alternatives if it is not quite right either.
Best for: Teachers and students who want Weka's algorithms with a more visual, interactive canvas.
Pricing:
- Free and open source (GPL v3 or later)
- No paid tier
Watch out for:
- Orange has fewer algorithms and filters than Weka's package manager offers, so niche methods from papers may be missing.
- It is desktop only. There is no browser version, so every student needs a local install.
Official site: Orange.
KNIME
KNIME Analytics Platform suits Weka users whose real problem is data preparation. It is a free, open source desktop app with a large library of nodes for joining, cleaning, reshaping and loading data from databases, files and APIs, plus Python and R scripting nodes. Weka's preprocess tab handles filters on one table; KNIME handles whole pipelines.
The paid tiers run workflows in KNIME's cloud. The KNIME Pro page lists €19 a month with 120 credits and 500 interactions with K-AI, the built-in assistant, and supports scheduled runs.
Best for: Analysts who spend more time cleaning and joining data than training models.
Pricing:
- Analytics Platform: free and open source
- Pro: €19 a month, 120 credits, 500 K-AI interactions; extra credits €0.025 each
- Team and Business Hub: priced by capacity, contact sales
Watch out for:
- Pro credits go quickly. The smallest executor burns 1 credit per minute, so 120 credits cover about two hours.
- The node library is large enough that beginners can get lost; Weka's Explorer is simpler for a first classifier.
Official site: KNIME.
scikit-learn
Most Weka algorithms have a close scikit-learn counterpart: J48 roughly maps to DecisionTreeClassifier, SMO to SVC, RandomForest to RandomForestClassifier. Moving to it means writing Python in a notebook instead of clicking, but you gain pipelines, cross-validation and model export that sit inside the same code as pandas and plotting.
PyPI lists version 1.9.1, released on September 10, 2026, under the BSD 3-Clause licence, with Python 3.11 or newer required. It is the library most data science courses and job ads expect.
Best for: Students and analysts ready to move from a GUI to Python code.
Pricing:
- Free and open source (BSD 3-Clause)
Watch out for:
- Deep learning is out of scope; the scikit-learn FAQ says the small neural network module gets bug fixes only and points to PyTorch or TensorFlow.
- There is no GUI at all. You need to be comfortable writing and debugging Python.
Official site: scikit-learn.
H2O-3
Weka loads the whole dataset into one Java heap, which is where many projects hit a wall. H2O-3 is an open source, distributed engine that spreads data and training across a cluster. It offers gradient boosting, random forests, GLMs and deep learning, plus AutoML that trains a batch of models and ranks them on a leaderboard.
You drive it from Python, R or a REST API. PyPI lists version 3.46.0.12, uploaded on August 12, 2026, under Apache 2.0.
Best for: Projects that outgrew a single machine's memory or want automated model comparison.
Pricing:
- H2O-3: free and open source (Apache 2.0)
- H2O.ai commercial platform: contact sales
Watch out for:
- Its web interface, H2O Flow, is deprecated and will be removed in a future release, so plan on code.
- Exporting POJO or MOJO models from Flow is blocked in the open source build.
Official site: H2O-3.
Dataiku
Dataiku is for the moment a Weka project becomes a team project. Analysts build visual recipes in a shared flow, data scientists add Python, R or SQL in the same project, and the platform handles scheduling, model deployment and permissions in the browser.
The get-started page offers a 14-day Dataiku Cloud trial with no credit card, and a Free Edition you install on your own Mac or Linux machine.
Best for: Companies moving models from one analyst's laptop to a governed shared platform.
Pricing:
- Dataiku Cloud: 14-day free trial, no card
- Free Edition: self-installed
- Paid plans: no public price; contact sales
Watch out for:
- Windows support for the Free Edition is labelled experimental.
- With no published prices, the cost after the trial stays unknown until sales quotes it.
Official site: Dataiku.
IBM SPSS Modeler
Some organisations need a supported, commercial version of what Weka does. IBM SPSS Modeler is a desktop visual modeller for Windows and Mac where you build streams of data preparation and modelling nodes, close in spirit to Weka's Knowledge Flow but with IBM support behind it.
It is one of the few tools in this group with a published price. The pricing page shows a subscription from USD 529 a month for one authorized user on a one-year term. Professional, Premium and Gold editions are sold as one-time purchases.
Best for: Universities and companies that already license SPSS and want a visual modeller with vendor support.
Pricing:
- Subscription: from USD 529 a month, one authorized user, one-year term
- Professional, Premium, Gold: one-time purchase, no public price
- Free trial offered
Watch out for:
- At USD 529 a month, one seat costs more than USD 6,000 a year for features Weka gives away.
- The perpetual editions have no published price, so budgeting needs a quote.
Official site: IBM SPSS Modeler.
RapidMiner AI Studio
RapidMiner used to be the default commercial step up from Weka. Its ownership changed twice: Altair bought RapidMiner in September 2022, and Siemens closed its purchase of Altair on March 26, 2025. The desktop tool is now sold as RapidMiner AI Studio, "part of Intelligence Center x".
The drag-and-drop canvas with Python and R still exists, and it deploys to any major cloud, on premises or hybrid. Siemens sells it by quote through its product page, with no public trial or free edition.
Best for: Enterprises that buy from Siemens and want a supported visual ML platform.
Pricing:
- No public price; contact sales
Watch out for:
- The free Studio edition that students once used is gone from the product page.
- An Altair community thread reports that the old download link redirects to a page with nothing to download.
Official site: RapidMiner AI Studio.
How to choose
| If you need | Pick |
|---|---|
| A free visual tool for a class | Orange |
| Heavy data cleaning and joins | KNIME |
| Python code that employers expect | scikit-learn |
| Data larger than one machine's memory | H2O-3 |
| A shared platform for a data team | Dataiku |
| A vendor contract with a list price | IBM SPSS Modeler |
| Siemens enterprise tooling | RapidMiner AI Studio |
When to stay with Weka
Stay with Weka if you teach or study classic machine learning. Its Explorer and Experimenter let you compare dozens of algorithms with proper statistics in a few clicks, and textbooks and courses still use it. It is free under the GPL and plugs into R and Python through packages, per its Wikipedia entry. Stay on the 3.8 branch for coursework, since 3.9 is the development line.
Frequently asked questions
Is Weka still maintained?
Yes. Weka 3.8.7 and 3.9.7 were released in late March 2026 and announced on April 2, 2026, with Java 25 support and new ARM builds for Windows and Linux. The download page says new releases normally come once or twice a year.
What is the Python equivalent of Weka?
scikit-learn. It covers the same classic algorithms, decision trees, SVMs, random forests, naive Bayes and clustering, as Python code. If you want a GUI that still runs on Python, Orange installs with pip or conda and adds a visual canvas.
Is Orange better than Weka?
For visual exploration and teaching, usually yes, because charts and models update live on one canvas. Weka has more algorithms and a stronger Experimenter for comparing them statistically.
Is there a free alternative to Weka for large datasets?
H2O-3. It is free under Apache 2.0 and spreads data and training across a cluster. KNIME Analytics Platform is also free, and its paid Pro and Hub tiers run workflows on KNIME's servers instead of your laptop.
