Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

spaCy

spacy.io

  • Free plan
  • Windows
  • macOS
  • Linux
  • Self-hosted
7.9 Club score Visit spacy.io

What spaCy is

spaCy is a free, open-source Python library for building natural language processing applications, with production use and large text workloads in view. Its components support named entity recognition, part-of-speech tagging, dependency parsing, sentence segmentation, text classification, lemmatization, morphology, and entity linking. The maker lists support for more than 75 languages and 84 trained pipelines for 25 languages, though trained pipelines are not available for every language. Developers can configure and train pipelines, describe training runs in configuration files, and package trained pipelines for installation as Python packages. spaCy can load pretrained transformers through spacy-transformers and work with custom models in PyTorch, TensorFlow, and other frameworks. spaCy Projects supports workflow tasks such as asset downloads, command execution, checksum verification, and caching, with integrations for tracking, demos, model serving, and distributed training. It runs on Linux, macOS, and Windows and can be installed through pip, conda, or source. It is a library for building applications, not a hosted web service. Transformer models require a GPU to run effectively.

Who it suits

spaCy suits developers building NLP applications, especially for production workloads with large volumes of text. It is not a fit for someone seeking a ready-made hosted web application.

Good to know, and what to watch

  • Free and open source.
  • Covers a broad set of NLP tasks.
  • Supports 75+ languages.
  • Custom pipelines can be packaged for installation.
  • Can run without data leaving the user's servers.
  • Trained pipelines are not available for every language.
  • Transformer models require a GPU to run effectively.
  • It does not provide a hosted web application.

spaCy

Free

Free plan

Getting started

  1. Visit the spaCy website.
  2. Install spaCy through pip, conda, or from source.
  3. Choose a CPU or GPU installation option.
  4. Configure or train a pipeline, or use a trained pipeline where available.
  5. Package trained pipelines as Python packages for installation.

Limits to check

Transformer models require a GPU to run effectively, while word vectors are computationally efficient. Trained pipelines are not available for every language.

The Club's review

spaCy is a production-oriented Python toolkit for teams building their own natural-language processing systems. It is best for developers who need control over models, deployment and data handling rather than a finished analysis application. The main choice is between its unusually broad NLP and workflow capabilities and the engineering work required to operate a library yourself.

Overview

Explosion, a Berlin-based AI developer-tools company, has maintained spaCy since 2015. The free, open-source library is built to process large text volumes and covers entity extraction and linking, text classification, part-of-speech tagging, dependency parsing, sentence segmentation, lemmatization and morphological analysis. That combination makes it suitable for production pipelines rather than occasional manual coding.

Language reach is substantial: the project supports more than 75 languages and provides 84 trained pipelines for 25 of them. Tokenization may cover additional languages, but a ready-trained pipeline is not guaranteed, so multilingual projects may require your own training work. spaCy runs on your servers without sending data away, and trained pipelines can be installed as Python packages with model data stored anywhere on the file system.

Key features

Pipeline design and training

You can assemble components, train custom pipelines and package the result for repeatable installation. Configuration files describe training runs, making experiments reproducible instead of dependent on undocumented settings. spaCy Projects extends that approach across a workflow: it can fetch source assets, execute commands, verify checksums and cache files through different backends.

Transformers and frameworks

The spacy-transformers integration loads pretrained transformer models and lets pipeline components share transformer embeddings, which is useful for multi-task systems. Custom models from PyTorch, TensorFlow and other frameworks can also be incorporated. Transformer workloads require a GPU to run effectively; conventional word vectors are much less demanding computationally, an important distinction when budgeting infrastructure.

Deployment and integrations

Project integrations cover data and experiment tracking, demonstrations, model serving, distributed training and uploading pipelines to Hugging Face Hub. Installation is available through pip, conda or source, with CPU and GPU options. CuPy supports GPU use, and a Visual Studio Code extension helps with configuration files. Community help is available through GitHub Discussions, the issue tracker and Stack Overflow.

Pricing

spaCy — free. This is a free, open-source library offered under AGPL or a commercial license, with no seat, usage or term charge stated. It gives developers the complete toolkit for building, training and deploying pipelines, but it does not include a hosted application or managed operations. Explosion also sells custom NLP consulting and implementation; those services use fees quoted up front rather than a published price list.

The free plan suits engineering teams able to provide their own Python environment, compute and maintenance. Organizations that need commercial licensing guidance or implementation help should consider the paid consulting route. There is no SaaS workspace, browser interface or vendor-managed quota to substitute for that in-house responsibility.

Platforms

spaCy runs on Linux, macOS and Windows, including Unix/Linux and OS X installations, and supports 64-bit CPython 3.7 and newer. It can be self-hosted, with CPU or GPU installation choices. On-premise execution and local data storage are strong fits for privacy-sensitive deployments, while transformer-heavy pipelines need suitable GPU hardware to perform effectively.

Who it's for

Choose spaCy if you are a Python developer, ML team or organization building repeatable NLP services at scale and need custom components, framework interoperability or control over where data and models live. It is also a strong foundation for teams managing experiments and deployment through code and project configuration.

Look elsewhere if your priority is a ready-made web application for qualitative analysis, collaborative review or nontechnical users. spaCy supplies the library and workflow building blocks; your team must create the surrounding product, interface and operations.

Pros and cons

  • Pro: Broad task coverage, including entity linking and morphology, supports complete text-processing pipelines in one library.
  • Pro: Support for 75+ languages and 84 trained pipelines across 25 languages gives multilingual projects a substantial starting point.
  • Pro: Self-hosting, local file storage and processing without data leaving your servers support strict data-location requirements.
  • Pro: Reproducible configuration-based training, packaging and spaCy Projects connect experimentation to deployment.
  • Con: Trained pipelines cover only 25 languages, so some language projects require custom model development.
  • Con: Transformer pipelines generally need a GPU for effective performance, adding infrastructure demands.
  • Con: There is no hosted application or managed service; teams must build and run the surrounding system themselves.

Alternatives

MAXQDA is a better fit for desktop qualitative analysis on macOS or Windows, with a paid product and a limited free AI Assist add-on. Transana suits users seeking a macOS or Windows qualitative-analysis application with a demo and trial rather than a Python library. Quirkos is worth choosing for cloud collaboration across web and desktop/mobile platforms, including a student plan from $5 a month billed annually. KH Coder offers a freemium desktop route, including a free limited edition and a $99 one-time Mac edition. QualCoder is another free, open-source option for qualitative data analysis on Linux, macOS and Windows. Condens is preferable when you need a hosted research workspace, with plans from €15 a month. Dovetail provides a web/API product with a free plan for one channel and one project. For a broader comparison, see Text Analytics Software.

Verdict

spaCy is a strong recommendation for developers and technical teams that need a free, extensible NLP foundation they can train, package and run on their own infrastructure. Its task breadth, language coverage and reproducible workflow tooling are the main reasons to choose it. Look elsewhere when you need a finished, hosted application, visual collaboration or vendor-managed operations instead of a library your team must engineer around.

The verdict

spaCy is a flexible library for developers who want to build and manage NLP pipelines. Its broad language and framework support comes with a distinction to note: it is a software library, not a hosted application.

Questions people ask

How much does spaCy cost?

The listed spaCy plan is free. The library is available under AGPL or a commercial license.

Is spaCy open source?

Yes. It is described as a free, open-source Python library and is available under AGPL or a commercial license.

Which operating systems does spaCy support?

It runs on Linux, macOS, and Windows, including self-hosted deployments.

Does spaCy offer a hosted web application?

No. spaCy is a library for building NLP applications, not a software-as-a-service product or web application.

Who makes spaCy?

Explosion identifies itself as spaCy's maker and describes itself as a software company specializing in developer tools and services for AI and natural language processing.

Are paid services available?

Explosion offers custom NLP consulting and implementation, with fees quoted up front.

What the maker publishes

Custom pipelines
Users can configure and train custom pipelines, package them as Python packages, and manage end-to-end workflows with spaCy Projects. source
Data privacy
Explosion says spaCy can run without data leaving the user's servers. source
Deployment
Trained pipelines can be installed as Python packages, and their data can be stored anywhere on the file system. source
Founded
2015 source
Framework integration
The homepage says spaCy supports custom models in PyTorch, TensorFlow and other frameworks. source
Hardware limit
The documentation says transformer models require a GPU to run effectively, while word vectors are computationally efficient. source
Headquarters
Berlin, Germany source
Installation
The documentation lists installation through pip, conda or source, with CPU and GPU options. source
Integrations
The documentation describes integrations with transformer pipelines, CuPy for GPU support and an extension for Visual Studio Code configuration files. source
Languages
The site lists support for 75+ languages and 84 trained pipelines for 25 languages. source
Language support
The maker’s homepage states that spaCy supports 75+ languages and offers 84 trained pipelines for 25 languages. source
Maker
Explosion identifies itself as the maker of spaCy and describes itself as a software company specializing in developer tools and services for AI and natural language processing. source
Model availability
The documentation notes that although spaCy supports tokenization for many languages, trained pipelines are not available for every language. source
NLP features
Its components include named entity recognition, part-of-speech tagging, dependency parsing, sentence segmentation, text classification, lemmatization and morphological analysis. source
NLP tasks
Its components cover named entity recognition, part-of-speech tagging, dependency parsing, sentence segmentation, text classification, lemmatization, morphology and entity linking. source
Operating systems
The installation documentation says spaCy runs on Unix/Linux, macOS/OS X and Windows and is compatible with 64-bit CPython 3.7+. source
Paid services
Explosion offers custom NLP consulting and implementation, with fees quoted up front; the page does not publish a price list. source
Product boundary
spaCy is a library for building NLP applications; it does not provide a software-as-a-service product or web application. source
Project integrations
spaCy Projects describes integrations for data and experiment tracking, demos, model serving, distributed training and uploading pipelines to Hugging Face Hub. source
Purpose
spaCy is a Python library for advanced natural language processing, designed for production use and processing large volumes of text. source
Security
The official pages opened for this research did not state a security certification or compliance standard. source
Support
The site links to GitHub Discussions, an issue tracker and Stack Overflow for community help. source
Training
Its training system uses configuration files to describe training runs and make experiments reproducible. source
Transformer integration
spaCy can load pretrained transformer models through spacy-transformers and supports sharing transformer embeddings across pipeline components. source

Checked against the maker's pages on 7 Oct 2026.