Skip to content
SPECIFICTION
Get listed

Detoxify

github.com

Free plan LinuxSelf-hosted
Visit github.com
Spec score 7.6 #12 of 22 in Content Moderation Tools Software

What it is

Detoxify is a free, self-hosted library that provides trained models and code to classify toxic comments. Its original, unbiased, and multilingual models are based on BERT, RoBERTa, and XLM-RoBERTa, respectively. Depending on the model, outputs include categories such as threats, obscenity, insults, identity attacks, and explicit sexual content. The multilingual model is intended for English, French, Spanish, Italian, Portuguese, Turkish, and Russian. The project describes research, fine-tuning on carefully constructed representative datasets, and helping moderators flag harmful content faster as intended uses. It can be installed with pip and called from Python to predict labels for one string or a list. Inference uses CPU by default and accepts a PyTorch device such as CUDA. The project warns that profanity-associated words may be labeled toxic regardless of tone or intent, potentially biasing results against vulnerable minority groups. Inference uses Transformers and PyTorch Lightning; training also requires the Kaggle API to download data. API access can be requested by email. The library is distributed under Apache-2.0.

Spec sheet

Purpose
Detoxify provides trained models and code to classify toxic comments from three Jigsaw challenges.Read 8 Oct 2026 · github.com
License
The project is distributed under the Apache-2.0 license.Read 8 Oct 2026 · github.com
API access
The repository says to email [email protected] to request access to its API.Read 8 Oct 2026 · github.com
Bias limitation
The README warns that profanity-associated words may be classified as toxic regardless of tone or intent, potentially creating bias toward vulnerable minority groups.Read 8 Oct 2026 · github.com
Categories
Model outputs include toxicity-related categories such as threats, obscenity, insults, identity attacks, and explicit sexual content, depending on the model.Read 8 Oct 2026 · github.com
Dependencies
Inference uses Hugging Face Transformers and PyTorch Lightning, while training also requires the Kaggle API to download data.Read 8 Oct 2026 · github.com
Install
The library can be installed with pip and used from Python to predict labels for one string or a list of strings.Read 8 Oct 2026 · github.com
Languages
The multilingual model is intended for English, French, Spanish, Italian, Portuguese, Turkish, and Russian.Read 8 Oct 2026 · github.com
Models
The library offers original, unbiased, and multilingual models based on BERT, RoBERTa, and XLM-RoBERTa, respectively.Read 8 Oct 2026 · github.com
Runtime
Inference supports CPU by default and accepts a PyTorch device such as CUDA.Read 8 Oct 2026 · github.com
Training data
The models were trained for the 2018, 2019, and 2020 Jigsaw toxic-comment challenges using Wikipedia Comments and Civil Comments data.Read 8 Oct 2026 · github.com
Use
The project describes its intended uses as research, fine-tuning on carefully constructed representative datasets, and helping moderators flag harmful content faster.Read 8 Oct 2026 · github.com

Plans and prices

Open-source library Free Trained models and code for toxic-comment classification; API access by email Read 8 Oct 2026 · github.com

Every Detoxify plan and price →

Strong points and limits

Strong points

  • Free, self-hosted library under Apache-2.0.
  • Offers three models based on BERT-family architectures.
  • Multilingual model covers seven listed languages.
  • Accepts one string or a list of strings.

Limits

  • Profanity may be labeled toxic regardless of context.
  • Image, audio, and video moderation are not supported.
  • Training requires the Kaggle API to download data.

Who it suits

It suits researchers and moderators who need self-hosted text classification, or teams fine-tuning models on representative datasets. Its listed scope does not extend to image, audio, or video moderation.

Verdict

Detoxify provides free, self-hosted text classification with several models and a multilingual option. Its documented bias limitation matters when interpreting labels, particularly for profanity-associated words.

Where Specifiction places it

Alternatives to DetoxifyAll alternatives →