thinc

A refreshing functional take on deep learning, compatible with your favorite libraries

thinc Permissive license MIT Active ★ 2,890 v9.1.1 released 2024-09-12

Install

thinc on PyPI

pip

pip install thinc

uv

uv add thinc

poetry

poetry add thinc

Package facts

License	MIT (permissive)
Python support	supports the current Python release (>=3.9)
Install friction	medium — platform-specific wheel
Runtime dependencies	12 — blis, murmurhash, cymem, preshed, wasabi, srsly, catalogue, confection, setuptools, numpy, pydantic, packaging
Maintenance	actively maintained — 700 days since the last release
Last repo commit	2026-03-27
First released	2014-10-02
Popularity	one of the top 1,000 most-downloaded packages on PyPI (30-day window, as of 2026-08-13)
Known vulnerabilities	none known (OSV.dev, checked 2026-08-13)

Evidence: thinc-9.1.1-cp310-cp310-macosx_10_9_x86_64.whl; thinc-9.1.1-cp310-cp310-macosx_11_0_arm64.whl; thinc-9.1.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl; thinc-9.1.1-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl; thinc-9.1.1-cp310-cp310-musllinux_1_2_i686.whl; thinc-9.1.1-cp310-cp310-musllinux_1_2_x86_64.whl; thinc-9.1.1-cp310-cp310-win_amd64.whl; thinc-9.1.1-cp311-cp311-macosx_10_9_x86_64.whl; thinc-9.1.1-cp311-cp311-macosx_11_0_arm64.whl; thinc-9.1.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl; thinc-9.1.1-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl; thinc-9.1.1-cp311-cp311-musllinux_1_2_i686.whl; thinc-9.1.1-cp311-cp311-musllinux_1_2_x86_64.whl; thinc-9.1.1-cp311-cp311-win_amd64.whl; thinc-9.1.1-cp312-cp312-macosx_10_9_x86_64.whl; thinc-9.1.1-cp312-cp312-macosx_11_0_arm64.whl; thinc-9.1.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl; thinc-9.1.1-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl; thinc-9.1.1-cp312-cp312-musllinux_1_2_i686.whl; thinc-9.1.1-cp312-cp312-musllinux_1_2_x86_64.whl

Development Status :: 5 - Production/StableEnvironment :: ConsoleIntended Audience :: DevelopersIntended Audience :: Science/ResearchLicense :: OSI Approved :: MIT LicenseOperating System :: MacOS :: MacOS XOperating System :: Microsoft :: WindowsOperating System :: POSIX :: LinuxProgramming Language :: CythonProgramming Language :: Python :: 3Programming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.9Topic :: Scientific/Engineering

About thinc

from the package's own PyPI description — quoted content, verbatim

Thinc: A refreshing functional take on deep learning, compatible with your favorite libraries

From the makers of spaCy and Prodigy

Thinc is a lightweight deep learning library that offers an elegant, type-checked, functional-programming API for composing models, with support for layers defined in other frameworks such as PyTorch, TensorFlow and MXNet. You can use Thinc as an interface layer, a standalone toolkit or a flexible way to develop new models. Previous versions of Thinc have been running quietly in production in thousands of companies, via both spaCy and Prodigy. We wrote the new version to let users compose, configure and deploy custom models built with their favorite framework.

tests (image) [![Current Release...

Read as markdown · JSON record · Source repository · Homepage

AI interpretation — verify before relying

AI-generated interpretation of the package facts above; every digit, version, license, or vulnerability id it cites is grounded in the facts already shown on this page

Thinc is a lightweight deep learning library offering a functional-programming API for composing neural network models with type-checking support and interoperability with PyTorch, TensorFlow, and MXNet.

Medium install friction due to 12 runtime dependencies including numpy, pydantic, and specialized packages like blis and cymem. Extensive wheel coverage across Python 3.9–3.12 and major platforms (macOS, Linux, Windows) mitigates friction. Active maintenance with last commit in March 2026 and 700 days since last release suggests stable, production-ready status.

MIT license (permissive) allows unrestricted use, modification, and distribution in both open-source and commercial projects with minimal obligations beyond attribution.

Usage

pip install thinc

import thinc.api as thinc
from thinc.api import chain, with_array, Model

# Define a simple model using functional composition
model = chain(thinc.layers.Relu(nO=64), thinc.layers.Softmax())

Requires Python 3.9 or newer; if PyTorch is installed alongside Python 3.7+, uninstall the dataclasses package to avoid incompatibility.

Verdict: Thinc is a production-ready, actively maintained deep learning framework well-suited for composing custom models with type safety and framework interoperability. Its MIT license and top-1000 popularity tier reflect broad adoption, particularly within the spaCy ecosystem. Medium install friction is offset by comprehensive wheel support and stable maintenance; no known security vulnerabilities.

Needs verification

Whether the 700-day gap since last release reflects a stable mature state or delayed maintenance cycle
GPU backend availability and any additional system dependencies beyond the listed runtime packages
Performance characteristics and typical use cases beyond the documented MNIST and NLP examples

functional deep learning libraryneural network composition frameworkpytorch tensorflow wrappertype-checked model buildinglightweight ml toolkitdeep learning apimodel configuration system