All Models

This model is designed for sentence similarity and feature extraction tasks. It supports various datasets including s2orc, flax-sentence-embeddings/stackexchange_xml, ms_marco, gooaq, yahoo_answers_topics, code_search_net, search_qa, eli5, snli, multi_nli, wikihow, natural_questions, trivia_qa, and several embedding data sources such as sentence-compression, flickr30k-captions, altlex, simple-wiki.

Image classification model licensed under Apache 2.0.

Electra is a model designed for natural language understanding tasks. It is trained to distinguish between real and fake text, making it useful for various applications in NLP.

Detects age group with about 59% accuracy based on an image. License: Apache 2.0. Metrics include accuracy and F1 score. Base model used is Google ViT Base Patch 16 224 in 21k. The model is part of the image classification pipeline and utilizes the dataset nateraw/fairface.

Bert is a language model trained on the BookCorpus and Wikipedia datasets. It is licensed under Apache 2.0 and is designed for various natural language processing tasks.

This model is designed for vision tasks. It includes candidate labels such as playing music and playing sports. An example title is Cat & Dog. You can find a sample image at https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-dog-music.png.

Image classification model using the timm library. It is licensed under Apache 2.0 and trained on the ImageNet-1K dataset.

Roberta Large is a language model trained on datasets including BookCorpus and Wikipedia. It is licensed under MIT and is designed for various natural language processing tasks.

License: Apache 2.0

This model is designed for sentence similarity and feature extraction using sentence-transformers. It supports various tasks such as text embeddings inference and is compatible with datasets including s2orc, flax-sentence-embeddings/stackexchange_xml, ms_marco, gooaq, yahoo_answers_topics, code_search_net, search_qa, eli5, snli, multi_nli, wikihow, natural_questions, trivia_qa, embedding-data/sentence-compression, embedding-data/flickr30k-captions, and embedding-data/altlex.

Segmentation is a model for speaker diarization, speaker change detection, speaker segmentation, voice activity detection, and overlapped speech detection. It is part of the pyannote audio framework, which aims to improve the understanding of audio and voice data. The model is open-source under the MIT license, and the collected information will help acquire a better knowledge of the pyannote.audio user base and assist its maintainers in further improvements.

Roberta Base is a language model trained on datasets including BookCorpus and Wikipedia. It is licensed under MIT.

Adetailer is a model that utilizes the Apache 2.0 license and is built on the Ultralytics library. It is designed to work with datasets such as WIDER FACE and skytnt/anime-segmentation, and it is tagged with PyTorch.

DistilBERT is a smaller, faster, cheaper, and lighter version of BERT. It is designed for natural language processing tasks and is trained on the BookCorpus and Wikipedia datasets. The model is licensed under Apache 2.0 and is tagged with exbert.

This model is designed for multilingual sentence similarity tasks, supporting a wide range of languages including Arabic, Bulgarian, Catalan, Czech, Danish, German, Greek, English, Spanish, Estonian, Persian, Finnish, French, Galician, Gujarati, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Italian, Japanese, Georgian, Korean, Kurdish, Lithuanian, Latvian, Macedonian, Mongolian, Marathi, Malay, Burmese, Norwegian Bokmål, Dutch, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Albanian, Serbian, Swedish, Thai, Turkish, Ukrainian, Urdu, and Vietnamese. It is licensed under Apache 2.0 and is part of the sentence-transformers library, suitable for feature extraction and sentence similarity tasks.

The model is licensed under MIT and is designed for text processing tasks. It includes a sample input sequence: MQIFVKTLTGKTITLEVEPS<mask>TIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG.

Language model for English.

Wespeaker is a model designed for audio processing, specifically for voice and speech tasks such as speaker recognition, speaker verification, speaker identification, and speaker embedding. It utilizes the Voxceleb dataset and is licensed under CC BY 4.0. The model is part of the pyannote audio framework.

Speaker diarization model for audio processing, capable of speaker change detection, voice activity detection, overlapped speech detection, and automatic speech recognition. Tags include pyannote, pyannote-audio, and pyannote-audio-pipeline. The model is licensed under MIT.

Gpt2 is a language model designed for natural language processing tasks.

This model is designed for vision tasks. It includes candidate labels such as playing music and playing sports. An example title is Cat & Dog. You can find a sample image at the following source: https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-dog-music.png.

A video LLM with an Apache 2.0 license.

Moshiko is a PyTorch model designed for efficient training and inference using BF16 precision.

XLM Roberta Base is a multilingual model supporting a wide range of languages including Afrikaans, Amharic, Arabic, Assamese, Azerbaijani, Belarusian, Bulgarian, Bengali, Brazilian Portuguese, Bosnian, Catalan, Czech, Welsh, Danish, German, Greek, English, Esperanto, Spanish, Estonian, Basque, Persian, Finnish, French, Frisian, Irish, Scottish Gaelic, Galician, Gujarati, Hausa, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Kannada, Korean, Kurdish, Kyrgyz, Latin, Lao, Lithuanian, Latvian, Malagasy, Macedonian, Malayalam, Mongolian, Marathi, Malay, Nepali, Dutch, Norwegian, Oromo, Oriya, Punjabi, Polish, Pashto, Portuguese, Romanian, Russian, Arabic, Sindhi, Sinhala, Slovak, Slovenian, Somali, Albanian, Serbian, Sundanese, Swedish, Swahili, Tamil, Telugu, Thai, Tagalog, Turkish, Uighur, Ukrainian, Urdu, Uzbek, Vietnamese, and Xhosa.

This model is designed for text generation. It is licensed under the Apache 2.0 license, and you can find the license details at https://huggingface.co/Qwen/Qwen2.5-7B-Instruct/blob/main/LICENSE. The model supports the English language and is tagged for chat applications, utilizing the transformers library.

Trained without supervision following the approach described in Towards Unsupervised Dense Information Retrieval with Contrastive Learning. The associated GitHub repository is available here https://github.com/facebookresearch/contriever.

This model is designed for image-text-to-text tasks. It is licensed under the qwen-research license, which can be found at https://huggingface.co/Qwen/Qwen2.5-VL-3B-Instruct/blob/main/LICENSE. The model supports the English language and is part of the multimodal category, utilizing the transformers library.

Text generation model based on Qwen/Qwen3-0.6B. It is licensed under the Apache 2.0 license, and the license can be found at https://huggingface.co/Qwen/Qwen3-0.6B/blob/main/LICENSE. The base model is Qwen/Qwen3-0.6B-Base.

Pipeline tag: sentence-similarity. Tags: sentence-transformers, feature-extraction, sentence-similarity. License: MIT.

ColBERT is a model designed for efficient retrieval in information retrieval tasks. It is licensed under MIT and supports the English language. The model is associated with the ColBERT project.

This model is a token classification model based on DistilBERT, fine-tuned for privacy-related tasks. It is licensed under CC BY-NC 4.0 and is built on the distilbert-base-uncased architecture. The model was generated from a trainer and has been evaluated using the seqeval metric. It is trained on datasets including ai4privacy/pii-masking-200k and Isotonic/pii-masking-200k.

This model is designed for text generation. It is licensed under the Apache 2.0 license, and you can find the license details at https://huggingface.co/Gensyn/Qwen2.5-0.5B-Instruct/blob/main/LICENSE. The model supports English language and is tagged with chat, rl-swarm, and gensyn. It utilizes the transformers library.

This model is designed for diffusion in a single file format, optimized for use with ComfyUI.

Time series forecasting model licensed under Apache 2.0. It is a pretrained foundation model designed for time series analysis and forecasting.

This model is designed for text classification tasks, specifically fine-tuned on the SST-2 dataset from the GLUE benchmark. It achieves an accuracy of 0.9106 on the validation split of the SST-2 dataset. The model is licensed under Apache 2.0.

The model is designed for tasks such as sentence transformation, feature extraction, and sentence similarity using transformers. It has been evaluated on the MTEB Amazon Counterfactual Classification dataset with the following results: accuracy of 73.79, average precision of 37.22, and F1 score of 68.09.

Combined and quantized models for WanVideo, originating from here. Tags include diffusion single file and comfyui base model with versions Wan-AI/Wan2.1-VACE-14B and Wan-AI/Wan2.1-VACE-1.3B.

BERT is a multilingual model that supports a wide range of languages including Afrikaans, Albanian, Arabic, Armenian, Asturian, Azerbaijani, Basque, Bavarian, Belarusian, Bengali, Bihari, Bosnian, Breton, Bulgarian, Burmese, Catalan, Cebuano, Chechen, Chinese, Chuvash, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hebrew, Hindi, Hungarian, Icelandic, Interlingua, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Kazakh, Kyrgyz, Korean, Latin, Latvian, Lithuanian, Lombard, Macedonian, Malagasy, Malay, Malayalam, Marathi, Mongolian, Minangkabau, Nepali, Newar, Norwegian Bokmål, Norwegian Nynorsk, Occitan, Persian, Piedmontese, Polish, Portuguese, Punjabi, Romanian, Russian, Scots, Serbian, Serbian (Cyrillic), Slovak, Slovenian, Spanish, Sundanese, Swahili, Swedish, Tagalog, Tajik, Thai, Tamil, Tatar, Telugu.

Feature extraction model for sentence similarity, supporting multilingual capabilities across various languages including Afrikaans, Amharic, Arabic, Assamese, Azerbaijani, Belarusian, Bulgarian, Bengali, Brazilian Portuguese, Bosnian, Catalan, Czech, Welsh, Danish, German, Greek, English, Esperanto, Spanish, Estonian, Basque, Persian, Finnish, French, Frisian, Irish, Scottish Gaelic, Galician, Gujarati, Hausa, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Kannada, Korean, Kurdish, Kyrgyz, Latin, Lao, Lithuanian, Latvian, Malagasy, Macedonian, Malayalam, Mongolian, Marathi, Malay, Burmese, Nepali, Dutch, Norwegian, Oromo, Oriya, Punjabi, Polish, Pashto, Portuguese, Romanian, Russian, Arabic, Sindhi, Sinhala, Slovak, Slovenian.

Automatic Speech Recognition model trained on the LibriSpeech dataset. It supports English language audio processing and is part of the Hugging Face ASR leaderboard. The model is licensed under Apache 2.0 and includes example audio samples for demonstration.

This model is designed for fill-mask tasks in the legal domain. It is trained on English language data and is licensed under CC BY-SA 4.0. The model can be used to predict masked words in legal texts, such as in the example: "The applicant submitted that her husband was subjected to treatment amounting to [MASK] whilst in the custody of police." A thumbnail image is available at https://i.ibb.co/p3kQ7Rw/Screenshot-2020-10-06-at-12-16-36-PM.png.

Automatic Speech Recognition model for the Russian language. It is licensed under Apache 2.0 and trained on datasets including Common Voice. The model is evaluated using metrics such as Word Error Rate (WER) and Character Error Rate (CER). It is tagged with audio, automatic-speech-recognition, and robust-speech-event, among others.

Text-to-speech model with a coqui public model license. License link: https://coqui.ai/cpml. Library name: coqui. Pipeline tag: text-to-speech. Example text: Once when I was six years old I saw a magnificent picture.

This model is designed for multilingual sentence similarity and feature extraction. It supports a wide range of languages including Arabic, Bulgarian, Catalan, Czech, Danish, German, Greek, English, Spanish, Estonian, Persian, Finnish, French, Galician, Gujarati, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Italian, Japanese, Georgian, Korean, Kurdish, Lithuanian, Latvian, Macedonian, Mongolian, Marathi, Malay, Burmese, Norwegian Bokmål, Dutch, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Albanian, Serbian, Swedish, Thai, Turkish, Ukrainian, Urdu, and Vietnamese. The model is licensed under Apache 2.0 and is part of the sentence-transformers library.

This model is designed for text ranking tasks. It is licensed under Apache 2.0 and is based on the cross-encoder/ms-marco-MiniLM-L12-v2 model. It utilizes the sentence-transformers library and is trained on the ms-marco dataset, specifically for the English language.

Llama 3.1 is a text generation model supporting multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. It is based on the Meta-Llama 3.1 8B architecture and is licensed under the Llama 3.1 Community License Agreement. The release date is July 23, 2024. The model is tagged with facebook, meta, pytorch, llama, and llama-3.

This model is designed for sentence similarity and feature extraction using transformers. It supports various datasets including s2orc, flax-sentence-embeddings/stackexchange_xml, ms_marco, gooaq, yahoo_answers_topics, code_search_net, search_qa, eli5, snli, multi_nli, wikihow, natural_questions, trivia_qa, and several embedding-data sources.

The model is designed for tasks such as sentence transformation, feature extraction, and sentence similarity using transformers. It is part of the MTEB model index, specifically for the Amazon Counterfactual Classification dataset in English. The model's configuration is set to English with a test split. The revision identifier is e8379541af4e31359cca9fbcf4b00f2671dba205. The performance metrics include an accuracy of 75.85%, an average precision of 38.57, and an F1 score of 69.69.

A model designed for diffusion in a single file format, optimized for use with ComfyUI.

This model is based on the Yi 1.5 34B architecture and is licensed under Apache 2.0. It was generated from a trainer and utilizes various datasets including cognitivecomputations/Dolphin-2.9, teknium/OpenHermes-2.5, m-a-p/CodeFeedback-Filtered-Instruction, cognitivecomputations/dolphin-coder, cognitivecomputations/samantha-data, microsoft/orca-math-word-problems-200k, Locutusque/function-calling-chatml, and internlm/Agent-FLAN.

This model is designed for image-text-to-text tasks. It supports the English language and is built using the transformers library under the Apache 2.0 license.

This model is designed for sentiment analysis on tweets. It is trained on the TweetEval dataset and is available under the CC BY 4.0 license.

BERT is a multilingual model that supports a wide range of languages including Afrikaans, Albanian, Arabic, Armenian, Asturian, Azerbaijani, Basque, Belarusian, Bengali, Bihari, Bosnian, Breton, Bulgarian, Burmese, Catalan, Cebuano, Chechen, Chinese, Chuvash, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, Galician, Georgian, German, Greek, Gujarati, Haitian Creole, Hebrew, Hindi, Hungarian, Icelandic, Interlingua, Indonesian, Irish, Italian, Japanese, Javanese, Kannada, Kazakh, Kyrgyz, Korean, Latin, Latvian, Lithuanian, Luxembourgish, Macedonian, Malagasy, Malay, Malayalam, Marathi, Minangkabau, Nepali, Newar, Norwegian Bokmål, Norwegian Nynorsk, Occitan, Persian, Piedmontese, Polish, Portuguese, Punjabi, Romanian, Russian, Scots, Serbian, Slovak, Slovenian, Spanish, Sundanese, Swahili, Swedish, Tagalog, Tajik, Tamil, Tatar, Telugu, Turkish, Ukrainian.

Text generation model licensed under Apache 2.0, utilizing the transformers library and tagged with vllm.

This model is designed for sentence similarity tasks and feature extraction using the sentence-transformers library. It is licensed under Apache 2.0.

The model is designed for sentence transformation, feature extraction, and sentence similarity tasks. It is part of the MTEB model index, specifically for the Amazon Counterfactual Classification dataset in English. The test split has a revision identifier of e8379541af4e31359cca9fbcf4b00f2671dba205. The model achieved an accuracy of 76.15%, an average precision of 39.32%, and an F1 score of 70.17%.

Gemma is a text generation model that requires users to review and agree to Google’s usage license. To access Gemma on Hugging Face, ensure you’re logged in and click the provided button. Requests are processed immediately.

This model is licensed under the Apache 2.0 license.

Dino is a vision model licensed under Apache 2.0.

This model is based on Qwen/Qwen3-0.6B-Base and is designed for tasks such as sentence similarity, feature extraction, and text embeddings inference. It is licensed under Apache 2.0 and utilizes transformers and sentence-transformers.

Kokoro is an open-weight TTS model with 82 million parameters. Despite its lightweight architecture, it delivers comparable quality to larger models while being significantly faster and more cost-efficient. With Apache-licensed weights, Kokoro can be deployed anywhere from production environments to personal projects.

This model is licensed under the Apache 2.0 license. It supports the English language and is designed for text generation tasks. The base model is Qwen/Qwen2.5-1.5B and it is compatible with the transformers library.

Time series forecasting model licensed under Apache 2.0. It is designed for time series analysis and includes pretrained models, foundation models, and time series foundation models.

The model supports a wide range of languages including but not limited to: Abkhazian, Afrikaans, Akan, Amharic, Arabic, Assamese, Avar, Aymara, Azerbaijani, Bashkir, Bambara, Belarusian, Bengali, Bislama, Tibetan, Bosnian, Bulgarian, Catalan, Czech, Chechen, Chuvash, Kurdish, Welsh, Danish, German, Divehi, Dzongkha, Greek, English, Esperanto, Estonian, Basque, Finnish, French, Western Frisian, Fula, Irish, Galician, Guarani, Gujarati, Chinese, Haitian Creole, Hausa, Hebrew, Hindi, Hungarian, Armenian, Igbo, Interlingua, Malay, Icelandic, Italian, Javanese, Japanese, Kannada, Georgian, Kazakh, Korean, Khmer, Kikuyu, Kinyarwanda, Kyrgyz, Korean, Komi, Lao, Latin, Latvian, Lingala, Lithuanian, Luxembourgish, Ganda, Marshallese, Malayalam, Marathi, Malay, Macedonian, Malagasy, Maltese, Mongolian, Maori, Burmese, Chinese, Dutch, Norwegian, Nepali, Nyanja, Occitan, Oromo, and others.

This model supports multiple languages including English, Chinese, German, Spanish, Russian, Korean, French, Japanese, Portuguese, Turkish, Polish, Catalan, Dutch, Arabic, Swedish, Italian, Indonesian, Hindi, Finnish, Vietnamese, Hebrew, Ukrainian, Greek, Malay, Czech, Romanian, Danish, Hungarian, Tamil, Norwegian, Thai, Urdu, Croatian, Bulgarian, Lithuanian, Latin, Maori, Malayalam, Welsh, Slovak, Telugu, Persian, Latvian, Bengali, Serbian, Azerbaijani, Slovenian, Kannada, Estonian, Macedonian, Breton, Basque, Icelandic, Armenian, Nepali, Mongolian, Bosnian, Kazakh, Albanian, Swahili, Galician, Marathi, Punjabi, Sinhala, Khmer, Shona, Yoruba, Somali, Afrikaans, Occitan, Georgian, Tajik, Sudanese, Gujarati, Amharic, Yiddish, Lao, Uzbek, Faroese, Haitian Creole, Pashto, Tatar, Norwegian Nynorsk, Maltese, Arabic (Saudi), Luxembourgish, Malay (Malaysian), Tibetan, Tagalog, Malagasy, Assamese, Tatar, Hawaiian, Lingala, Hausa, Bashkir.

Text generation model designed for English language inference.

This model supports multiple languages including English, Chinese, German, Spanish, Russian, Korean, French, Japanese, Portuguese, Turkish, Polish, Catalan, Dutch, Arabic, Swedish, Italian, Indonesian, Hindi, Finnish, Vietnamese, Hebrew, Ukrainian, Greek, Malay, Czech, Romanian, Danish, Hungarian, Tamil, Norwegian, Thai, Urdu, Croatian, Bulgarian, Lithuanian, Latin, Maori, Malayalam, Welsh, Slovak, Telugu, Persian, Latvian, Bengali, Serbian, Azerbaijani, Slovenian, Kannada, Estonian, Macedonian, Brazilian Portuguese, Basque, Icelandic, Armenian, Nepali, Mongolian, Bosnian, Kazakh, Albanian, Swahili, Galician, Marathi, Punjabi, Sinhala, Khmer, Shona, Yoruba, Somali, Afrikaans, Occitan, Georgian, Tajik, Sudanese, Gujarati, Amharic, Yiddish, Lao, Uzbek, Faroese, Haitian Creole, Pashto, Tatar, Nynorsk, Maltese, Arabic, and more.

A chatbot that can help code. License: Apache-2.0. Datasets include Cerebras SlimPajama-627B, BigCode Starcoderdata, HuggingFaceH4 Ultrachat 200k, and HuggingFaceH4 Ultrafeedback Binarized. The example widget demonstrates a Python function to calculate the first 10 digits of the Fibonacci sequence and print it to the CLI.

Tiny Qwen 2 For Causal LM is a model that utilizes the transformers library and is tagged with trl.

This model is designed for text generation. It is part of the Qwen library and is licensed under the Apache 2.0 license. The license can be found at https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507/blob/main/LICENSE.

A model designed for vision tasks, licensed under Apache 2.0.

Automatic Speech Recognition model trained on Portuguese language using datasets such as Common Voice. It is licensed under Apache 2.0 and includes metrics like WER and CER. The model is tagged with audio, automatic-speech-recognition, and robust-speech-event, among others.

This model is designed for automatic speech recognition in Japanese, utilizing the Wav2Vec2 architecture. It is trained on the Common Voice dataset and evaluates performance using Word Error Rate (WER) and Character Error Rate (CER) metrics. The model is tagged for audio processing and fine-tuning in speech recognition tasks.

This model is designed for text generation and supports multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. It utilizes the transformers library and is built on PyTorch. The model is tagged with keywords such as facebook, meta, pytorch, llama, and llama-3. It is licensed under llama3.2 and includes an extra gated prompt.

This model is designed for sentence similarity tasks, providing feature extraction and text embeddings inference. It supports various datasets including Stack Exchange, MS MARCO, and Yahoo Answers Topics, among others.

This model is designed for sentence similarity tasks and is based on the MTEB Amazon Counterfactual Classification dataset. It utilizes the transformers library and is tagged with sentence-transformers, gte, mteb, and transformers.js. The model is licensed under Apache 2.0 and supports the English language. The evaluation results show an accuracy of 73.01% and an average precision of 35.05% on the test split.

Text generation model licensed under Apache 2.0, utilizing the transformers library and tagged with vllm.

Text generation model with a focus on chat applications. It is licensed under an other license. The model is based on Qwen/Qwen2.5-3B and is compatible with the transformers library. The primary language supported is English.

BloomZ 560M is a text generation model trained on the bigscience/xP3 dataset. It supports multiple languages including ak, ar, as, bm, bn, ca, code, en, es, eu, fon, fr, gu, hi, id, ig, ki, kn, lg, ln, ml, mr, ne, nso, ny, or, pa, pt, rn, rw, sn, st, sw, ta, te, tn, ts, tum, tw, ur, vi, wo, xh, yo, zh, and zu. The model is compatible with various programming languages such as C, C++, C#, Go, Java, JavaScript, Lua, PHP, Python, Ruby, Rust, Scala, and TypeScript. It is licensed under bigscience-bloom-rail-1.0.

A model designed for image classification, utilizing the timm library. It is trained on the ImageNet-1K dataset and is suitable for various image classification tasks.

Pipeline tag: text-to-image inference: false.

Esmfold V1 is a model designed for protein structure prediction, leveraging advanced machine learning techniques to provide accurate and efficient results.

Fill-mask model for English language processing with MIT license.

Object detection model for signature detection using the Ultralytics YOLOv8 framework. It is based on the AGPL-3.0 license and utilizes the tech4humans/signature-detection dataset. The model evaluates performance using metrics such as F1 score, precision, and recall. It is implemented in PyTorch with library version 8.0.239.

Image classification model licensed under Apache 2.0, utilizing the timm library and supporting transformers.

Bert is a language model trained on the BookCorpus and Wikipedia datasets. It is licensed under Apache 2.0 and is tagged with exbert.

This model is designed for text classification and supports multilingual capabilities. It is licensed under Apache 2.0 and utilizes transformers, sentence-transformers, and text embeddings for inference.

This model is designed for time series forecasting. It is licensed under CC BY-NC 4.0 and is categorized as a pretrained foundation model for time series applications.

Trained for 1,000,000 steps with a batch size of 32 on Python code from the codeparrot/github-code-clean dataset, on the masked-language-modeling task.

Zero-shot classification model trained on the MultiNLI dataset. It is licensed under MIT. For more information, visit the thumbnail at https://huggingface.co/front/thumbnails/facebook.png.

A model designed for vision and language tasks in the fashion and ecommerce domain. It supports English and utilizes the transformers library. The model can classify candidate labels such as black shoe and red shoe, and includes an example of a black shoe. A widget is available for integration.

Token classification model for part-of-speech tagging using the Indonesian RoBERTa base architecture. It is based on the flax-community/indonesian-roberta-base model and has been trained on the indonlu dataset. The model evaluates performance using metrics such as precision, recall, F1 score, and accuracy. The results for the test split of the indonlu dataset show a precision of 0.9625 and a recall of 0.9625.

Language support includes English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. The model utilizes the transformers library and is designed for text generation. It is associated with tags such as facebook, meta, pytorch, llama, and llama-3. The license for this model is llama3.2 and it includes an extra gated prompt.

Various models to be used in llama.cpp CI workflow.

Vision model trained on the ImageNet-21k dataset. License: Apache-2.0. Inference capability: false.

A vision model designed for image classification tasks. It can categorize images into candidate labels such as playing music and playing sports. An example includes a cat and dog image.

DNABERT S is designed for biological and genomic applications, providing advanced capabilities for DNA sequence analysis.

This model is designed for image classification tasks using candidate labels such as playing music and playing sports. It includes a sample image of a cat and dog. The model index contains the name clip-vit-large-patch14-336 with no results listed.

This model is designed for summarization tasks, specifically trained on the CNN/DailyMail dataset. It achieves high performance with metrics such as ROUGE-1 at 42.95, ROUGE-2 at 20.81, and ROUGE-L at 30.62. The model is licensed under MIT and is suitable for generating concise summaries from longer texts.

This model is designed for image-text-to-text tasks and is licensed under the Apache 2.0 license.

This model is a fine-tuned version of Mistral designed for instruction-based tasks. It is built using the transformers library and is licensed under Apache 2.0. The model includes tags such as finetuned and mistral-common. The current version is 0.2, with a new version available as 0.3. Inference is not enabled. For more information on how personal data is processed, please refer to the Privacy Policy.

This model is designed for sentence similarity and feature extraction using the sentence-transformers library. It was generated from a trainer with a dataset size of 45 and utilizes the Multiple Negatives Ranking Loss. The base model is BAAI/bge-base-en-v1.5 and is tagged for sentence-transformers.

Image-text-to-text model licensed under Apache 2.0. It is based on the pretrained model OpenGVLab/InternVL3_5-241B-Pretrained and is fine-tuned on datasets including OpenGVLab/MMPR-v1.2 and OpenGVLab/MMPR-Tiny. The model supports multilingual capabilities and is tagged with internvl and custom_code.

This model is designed for vision tasks, specifically image classification. It is trained on datasets including ImageNet-1K and ImageNet-21K. Example images include a tiger, a teapot, and a palace.

This model is designed for sentence similarity tasks and supports multiple languages including Arabic, Bulgarian, Catalan, Czech, Danish, German, Greek, English, Spanish, Estonian, Persian, Finnish, French, Galician, Gujarati, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Italian, Japanese, Georgian, Korean, Kurdish, Lithuanian, Latvian, Macedonian, Mongolian, Marathi, Malay, Burmese, Norwegian Bokmål, Dutch, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Albanian, Serbian, Swedish, Thai, Turkish, Ukrainian, Urdu, and Vietnamese. It is licensed under Apache 2.0 and is part of the sentence-transformers library, focusing on feature extraction and sentence similarity.

ResNet-18 model for image classification, pretrained on ImageNet-1k dataset.

This model is designed for zero-shot image classification. It supports candidate labels such as playing music and playing sports. The model is available under the MIT license and includes a widget for displaying sample images.

Text generation model based on Qwen/Qwen3-8B-Base. Licensed under Apache 2.0. License details can be found at https://huggingface.co/Qwen/Qwen3-8B/blob/main/LICENSE.

This model is designed for recognizing and interpreting facial expressions.

A keypoint detection model designed for human pose estimation, built using the transformers library. It is licensed under Apache 2.0 and supports the English language.

This model is designed for embedding tasks using the MTEB framework. It includes various classification tasks with datasets such as MTEB Amazon Counterfactual Classification and MTEB Amazon Polarity. The model has achieved an accuracy of 75.04%, an average precision of 37.74%, and an F1 score of 68.93%. The configurations are set for the English language and the test split.

T5 Small is a multilingual model that supports English, French, Romanian, and German. It is designed for tasks such as summarization and translation, and is licensed under Apache 2.0. The model is trained on the C4 dataset.

--- license: bsd-3-clause ---

YOLO World object detection model.

This model is designed for multilingual sentiment analysis on Twitter data. It can understand and analyze sentiments expressed in various languages, making it useful for applications that require insights from diverse linguistic sources.

DeBERTa V3 Base is a language model designed for fill-mask tasks. It utilizes advanced techniques to improve performance on various natural language processing tasks.

Finbert is designed for financial sentiment analysis, providing insights into market trends and sentiments. It analyzes text data to determine the sentiment surrounding financial topics, such as stock movements and currency fluctuations.

Trellis is a library designed for image-to-3D processing. It provides tools and pipelines to convert 2D images into 3D representations, enabling various applications in computer vision and graphics.

XLM-RoBERTa is a multilingual model designed for various language tasks. It supports a wide range of languages including Afrikaans, Amharic, Arabic, Assamese, Azerbaijani, Belarusian, Bulgarian, Bengali, Brazilian Portuguese, Bosnian, Catalan, Czech, Welsh, Danish, German, Greek, English, Esperanto, Spanish, Estonian, Basque, Persian, Finnish, French, Frisian, Irish, Scottish Gaelic, Galician, Gujarati, Hausa, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Kannada, Korean, Kurdish, Kyrgyz, Latin, Lao, Lithuanian, Latvian, Malagasy, Macedonian, Malayalam, Mongolian, Marathi, Malay, Nepali, Dutch, Norwegian, Oromo, Odia, Punjabi, Polish, Pashto, Portuguese, Romanian, Russian, Arabic, Sindhi, Sinhala, Slovak, Slovenian, Somali, Albanian, Serbian, Sundanese, Swedish, Swahili, Tamil, Telugu, Thai, Tagalog, Turkish, Uyghur, Ukrainian, Urdu, Uzbek, Vietnamese, and Xhosa.

Whisper Small is a multilingual speech recognition model that supports a wide range of languages including English, Chinese, German, Spanish, Russian, Korean, French, Japanese, Portuguese, Turkish, Polish, and many more. It is designed to transcribe audio into text accurately across various languages, making it a versatile tool for applications in transcription, translation, and accessibility.

OpenRail license. Tags: text-to-image, stable-diffusion.

Stable Diffusion is a text-to-image model that allows for image generation based on textual descriptions. It is licensed under CreativeML OpenRail and includes tags such as stable-diffusion, stable-diffusion-diffusers.

Vision Transformer with GPT-2 for automatic image captioning, generating natural language descriptions of images.

Image captioning model that generates descriptive text for images. It supports English language and is licensed under the BSD-3-Clause.

Wav2Vec2 model for automatic speech recognition in Dutch, fine-tuned on Common Voice dataset.

This model is designed for vision tasks. It is licensed under Apache 2.0. The candidate labels include playing music and playing sports. An example title is Cat & Dog. You can find sample images at the following source: https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-dog-music.png.

3B parameter model based on Llama 3.2 3B Instruct.

Language model for English with tags including exbert.

Multilingual E5 Large is a model designed for sentence similarity and feature extraction tasks. It is part of the MTEB benchmark, specifically for the Amazon Counterfactual Classification dataset in English. The model has been evaluated on the test split with a revision identifier e8379541af4e31359cca9fbcf4b00f2671dba205. The performance metrics include an accuracy of 79.06%, an average precision of 43.49%, and an F1 score of 73.33%.

DeepSeek R1 Distill Qwen 32B is a model designed for advanced AI tasks, utilizing the transformers library under the MIT license.

This model is designed for image-text-to-text tasks. It supports the English language and is based on the Qwen/Qwen2-VL-2B architecture. It is licensed under Apache 2.0 and utilizes the transformers library.

Detects multiple languages including Arabic, Bulgarian, German, Greek, English, Spanish, French, Hindi, Italian, Japanese, Dutch, Polish, Portuguese, Russian, Swahili, Thai, Turkish, Urdu, Vietnamese, and Chinese. The model is based on XLM-Roberta and is licensed under MIT. It has been trained on the papluca/language-identification dataset and evaluated using metrics such as accuracy and F1 score.

A model for image segmentation licensed under Apache 2.0.

Neural sparse encoding model for OpenSearch, optimized for document retrieval and passage ranking using learned sparse representations.

--- license: mit ---

CLIP-based safety checker for Stable Diffusion, designed to filter inappropriate content.

DeepSeek OCR is a multilingual image-to-text model designed for optical character recognition. It supports various languages and is equipped with advanced vision-language capabilities. The model is built using the Transformers library and is available under the MIT license.

This model is designed for image-text-to-text tasks, supporting the English language. It utilizes the transformers library and is tagged as multimodal.

Wav2Vec2 model for automatic speech recognition in Arabic, trained on Common Voice and Arabic Speech Corpus datasets.

--- license: cc-by-nc-sa-4.0 ---

Wav2Vec2 model for automatic speech recognition in Chinese (Mandarin), trained on Common Voice dataset.

Qwen3-style causal language model with 1.7B parameters.

This model is designed for image-text-to-text tasks. It supports the English language and is built on the Qwen/Qwen2-VL-7B architecture. The model is licensed under Apache 2.0 and is compatible with the transformers library.

Automatic speech recognition model using the Transducer architecture. It is built with the NeMo library and supports English language. The model is trained on datasets such as nvidia/Granary and nvidia/nemo-asr-set-3.0. It features tags like FastConformer, Conformer, and is part of the hf-asr-leaderboard. Example audio samples include Librispeech sample 1 and Librispeech sample 2.

This model is designed for text generation and supports multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. It utilizes the transformers library and is built on the PyTorch framework. The model is tagged with keywords such as Facebook, Meta, and Llama, indicating its origin and functionality.

600M parameter RNN-T speech recognition model trained on multiple English speech datasets including LibriSpeech, Fisher, Switchboard, and Common Voice.

Depth Anything V2 Base model for monocular depth estimation.

Transformer model for recognizing and analyzing table structures in documents.

Image-text-to-text model licensed under Apache 2.0. The new version is Moondream 3 Preview.

--- datasets: - conceptofmind/cot_submix_original - conceptofmind/flan2021_submix_original - conceptofmind/t0_submix_original - conceptofmind/niv2_submix_original language: - en pipeline_tag: text-generation ---

Segmentation is a model designed for audio processing tasks such as speaker segmentation, voice activity detection, and overlapped speech detection. It is part of the pyannote audio framework, which aims to enhance the understanding of audio data. The model is licensed under MIT and is intended for research purposes. Users are encouraged to cite relevant papers in their publications to support the development and improvement of the pyannote.audio user base.

Llama 3.2 is a text generation model that supports multiple languages including English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai. It is built on the PyTorch library and is part of the Llama series developed by Meta. The model is designed for various text generation tasks and is tagged for easy discovery.

This model is designed for token classification tasks using the CoNLL2003 dataset. It achieves an accuracy of 0.9118, with a precision of 0.9212, recall of 0.9306, and an F1 score of 0.9259. The model is licensed under MIT.

This model is designed for sentence similarity tasks. It supports English language and is part of the sentence-transformers library. Key features include feature extraction and sentence similarity. It has been trained on various datasets including flax-sentence-embeddings/stackexchange_xml, ms_marco, gooaq, yahoo_answers_topics, search_qa, eli5, natural_questions, trivia_qa, embedding-data/QQP, embedding-data/PAQ_pairs, embedding-data/Amazon-QA, and embedding-data/WikiAnswers. The model is tagged for sentence similarity.

--- license: apache-2.0 language: en ---

Language model for text generation. Tags include facebook, meta, pytorch, llama, and llama-3. License is llama3. New version available: Llama 3.1 8B. Extra gated prompt is included.

This model is designed for text generation tasks. It is licensed under the Apache 2.0 license, and the license details can be found at the provided link. The model supports the English language and is compatible with the transformers library.

The model is designed for sentence similarity tasks using the sentence-transformers library. It supports feature extraction and is part of the MTEB benchmark, specifically for the Amazon Counterfactual Classification dataset in English. The model's performance metrics include an accuracy of 75.21%, an average precision of 38.58%, and an F1 score.

--- library_name: transformers tags: [] ---

This model is designed for speaker verification and identification using embeddings. It utilizes the ECAPA-TDNN architecture and is implemented in PyTorch. The model is trained on the VoxCeleb dataset and is licensed under Apache 2.0. It supports various applications in speech processing.

Language model licensed under Apache 2.0.

Multilingual E5 Base is designed for sentence similarity tasks and is part of the MTEB benchmark. It has been evaluated on the Amazon Counterfactual Classification dataset, achieving an accuracy of 78.97%, an average precision of 43.69%, and an F1 score of 73.38%. This model is suitable for various multilingual applications in natural language processing.

A diffusion model designed for use with ComfyUI, offering a streamlined interface for image generation.

--- language: fa datasets: - common_voice metrics: - wer - cer tags: - audio - automatic-speech-recognition - speech - xlsr-fine-tuning-week license: apache-2.0 model-index: - name: XLSR Wav2Vec2 Persian by Jonatas Grosman results: - task: name: Speech Recognition type: automatic-speech-recognition dataset: name: Common Voice fa type: common_voice args: fa metrics: - name: Test WER type: wer value: 30.12 - name: Test CER type: cer value: 7.37 ---

--- language: en tags: - exbert

This model is designed for text generation tasks in English. It is based on the Qwen 2.5 architecture and is compatible with the transformers library. The model is licensed under the Apache 2.0 license.

--- language: - es library_name: pysentimiento pipeline_tag: text-classification tags: - twitter - sentiment-analysis

--- license: apache-2.0 datasets: - sentence-transformers/msmarco language: - en base_model: - nreimers/BERT-Tiny_L-2_H-128_A-2 pipeline_tag: text-ranking library_name: sentence-transformers tags: - transformers ---

This model is designed for text generation. It is part of the transformers library and is licensed under the Apache 2.0 license. For more information, you can visit the license link at https://huggingface.co/Qwen/Qwen3-32B/blob/main/LICENSE.

--- license: apache-2.0 tags: - dino - vision ---

--- license: apache-2.0 language: - en library_name: transformers ---

--- language: fa ---

This model supports multiple languages, including Afrikaans, Amharic, Arabic, Assamese, Azerbaijani, Belarusian, Bulgarian, Bengali, Breton, Bosnian, Catalan, Czech, Welsh, Danish, German, Greek, English, Esperanto, Spanish, Estonian, Basque, Persian, Finnish, French, Frisian, Irish, Scottish Gaelic, Galician, Gujarati, Hausa, Hebrew, Hindi, Croatian, Hungarian, Armenian, Indonesian, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Kannada, Korean, Kurdish, Kyrgyz, Latin, Lao, Lithuanian, Latvian, Malagasy, Macedonian, Malayalam, Mongolian, Marathi, Malay, Nepali, Dutch, Norwegian, Oromo, Punjabi, Polish, Pashto, Portuguese, Romanian, Russian, Arabic, Sindhi, Sinhala, Slovak, Slovenian, Somali, Albanian, Serbian, Sundanese, Swedish, Swahili, Tamil, Telugu, Thai, Tagalog, Turkish, Uyghur, Ukrainian, Urdu, Uzbek, Vietnamese, Xhosa, Yiddish, and Chinese.

--- license: apache-2.0 tags: - dino - vision datasets: - imagenet-1k ---

Flux is a text-to-image model designed for image generation. It is licensed under the Flux-1-Dev Non-Commercial License. By clicking 'Agree', you agree to the FluxDev Non-Commercial License Agreement and acknowledge the Acceptable Use Policy.

This model is designed for text generation and supports English and French languages. It is tagged for natural language processing and code inference. The model has parameters such as temperature set to 0. It can assist users with various tasks, including providing ways to eat combinations of bananas and dragonfruits.

Base model for text-to-speech using the SpeechT5 architecture from Microsoft, implemented in the transformers.js library.

--- license: apache-2.0 language: - en tags: - biomedical - lexical semantics - bionlp - biology - science - embedding - entity linking --- ---

Bert Base Chinese is a language model designed for understanding and processing Chinese text. It is licensed under the Apache 2.0 license.

--- language: en license: apache-2.0 datasets: - bookcorpus - wikipedia ---

--- license: apache-2.0 datasets: - sentence-transformers/msmarco language: - en base_model: - cross-encoder/ms-marco-MiniLM-L12-v2 pipeline_tag: text-ranking library_name: sentence-transformers tags: - transformers ---

--- inference: true tags: - musicgen license: cc-by-nc-4.0 pipeline_tag: text-to-audio widget: - text: a funky house with 80s hip hop vibes example_title: Prompt 1 - text: a chill song with influences from lofi, chillstep and downtempo example_title: Prompt 2 - text: a catchy beat for a podcast intro example_title: Prompt 3 ---

MobileVit Small is a vision model designed for image classification. It is trained on the ImageNet-1K dataset. Example images include a tiger, a teapot, and a palace.

E5 Base V2 is a sentence-transformers model designed for sentence similarity tasks. It has been evaluated on the MTEB Amazon Counterfactual Classification dataset with a test split in English. The model achieved an accuracy of 77.78%, an average precision of 42.05%, and an F1 score of 72.12%. The revision used for this evaluation is e8379541af4e31359cca9fbcf4b00f2671dba205.

Grounding Dino Base is a model designed for zero-shot object detection. It operates under the Apache 2.0 license and is focused on vision tasks.

--- license: mit widget: - src: https://www.invoicesimple.com/wp-content/uploads/2018/06/Sample-Invoice-printable.png example_title: Invoice ---

Gemma is a model designed for image-to-text tasks, utilizing the transformers library. To access Gemma on Hugging Face, you must review and agree to Google's usage license. Ensure you are logged in to Hugging Face to proceed with the request.

A multilingual model designed for sentence similarity and text embeddings inference. It supports a wide range of languages including Afrikaans, Arabic, Azerbaijani, Belarusian, Bulgarian, Bengali, Catalan, Cebuano, Czech, Welsh, Danish, German, Greek, English, Spanish, Estonian, Basque, Persian, Finnish, French, Galician, Gujarati, Hebrew, Hindi, Croatian, Haitian Creole, Hungarian, Armenian, Indonesian, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Kannada, Korean, Kyrgyz, Lao, Lithuanian, Latvian, Macedonian, Malayalam, Mongolian, Marathi, Malay, Burmese, Nepali, Dutch, Norwegian, Punjabi, Polish, Portuguese, Quechua, Romanian, Russian, Sinhala, Slovak, Slovenian, Somali, Albanian, Serbian, Swedish, Swahili, Tamil, Telugu, Thai, Tagalog.

Time series forecasting model licensed under Apache 2.0. It is a pretrained foundation model designed for time series analysis. The latest version is amazon/chronos-2.

Voice activity detection identifies segments of audio that contain speech. It is useful for various applications in audio processing, including automatic speech recognition and speaker identification. The model is associated with tags such as pyannote, pyannote-audio, and voice-activity-detection. It has been trained on datasets like AMI, DiHard, and VoxConverse. The model is licensed under MIT. The collected information will help acquire a better knowledge of the pyannote.audio user base and assist maintainers in applying for grants to improve it further. Academic researchers are encouraged to cite relevant papers in their publications.

--- license: apache-2.0 language: - en tags: - ColBERT - RAGatouille - passage-retrieval ---

The model is designed for translation tasks, supporting languages such as English, French, Romanian, and German. It is trained on the C4 dataset and is tagged for summarization and translation. The model is licensed under Apache 2.0.

--- language: en license: mit ---

--- library_name: UniDepth tags: - model_hub_mixin - monocular-metric-depth-estimation - pytorch_model_hub_mixin ---

Automatic speech recognition model supporting English language. Licensed under MIT. Utilizes the transformers library. Tags include audio, automatic-speech-recognition, and transformers.js. Example audio samples include LibriSpeech sample 1 and LibriSpeech sample 2, available at https://cdn-media.huggingface.co/speech_samples/sample1.flac and https://cdn-media.huggingface.co/speech_samples/sample2.flac respectively. The model is designed for efficient and accurate transcription of spoken language.