Top Hugging Face models for NLP, vision, and audio tasks — links, descriptions, and demos included.

A curated list of the most powerful and popular models available on Hugging Face Hub, organized by task and domain. Each model includes direct links, detailed descriptions, and information about available demos.

• Natural Language Processing
  • Large Language Models
  • Text Classification
  • Named Entity Recognition
  • Question Answering
  • Text Summarization
  • Translation
  • Code Generation
• Computer Vision
  • Image Classification
  • Object Detection
  • Image Segmentation
  • Image Generation
  • Optical Character Recognition
• Audio Processing
  • Speech Recognition
  • Text-to-Speech
  • Audio Classification
  • Music Generation
• Multimodal
  • Vision-Language
  • Audio-Text
• Contributing
• License

The top models by downloads and community adoption:

• bert-base-uncased - The foundational BERT model
• gpt2 - OpenAI's GPT-2 model
• distilbert-base-uncased - Distilled BERT for efficiency
• openai/clip-vit-base-patch32 - CLIP for vision-language tasks

Cutting-edge models pushing the boundaries:

• openai/whisper-large-v2 - Best speech recognition
• runwayml/stable-diffusion-v1-5 - Leading image generation
• facebook/musicgen-large - Advanced music generation
• mistralai/Mixtral-8x7B-Instruct-v0.1 - Top instruction-following LLM

# Text Classification
from transformers import pipeline
classifier = pipeline("sentiment-analysis", 
                     model="cardiffnlp/twitter-roberta-base-sentiment-latest")
result = classifier("I love this awesome repository!")

# Image Classification  
classifier = pipeline("image-classification", 
                     model="google/vit-base-patch16-224")
result = classifier("path/to/image.jpg")

# Speech Recognition
transcriber = pipeline("automatic-speech-recognition",
                      model="openai/whisper-large-v2")
result = transcriber("path/to/audio.wav")

# Question Answering
from transformers import AutoTokenizer, AutoModelForQuestionAnswering
tokenizer = AutoTokenizer.from_pretrained("deepset/roberta-base-squad2")
model = AutoModelForQuestionAnswering.from_pretrained("deepset/roberta-base-squad2")

# Image Captioning
from transformers import BlipProcessor, BlipForConditionalGeneration
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")

• Text Classification: Start with RoBERTa-based models for accuracy
• Generation: Use GPT-style models (GPT-J, BLOOM, Llama 2)
• Question Answering: BERT/RoBERTa fine-tuned on SQuAD
• Multilingual: XLM-R, mBERT, or BLOOM for cross-lingual tasks

• Image Classification: ViT for state-of-the-art, ResNet for efficiency
• Object Detection: DETR for end-to-end, YOLO for speed
• Image Generation: Stable Diffusion for quality and control

• Speech Recognition: Whisper for multilingual, Wav2Vec2 for English
• TTS: SpeechT5 for quality, Tacotron for customization
• Audio Classification: AST for general audio understanding

• GLUE: General Language Understanding benchmark
• SuperGLUE: More challenging language understanding tasks
• SQuAD: Reading comprehension benchmark
• HellaSwag: Commonsense reasoning

• ImageNet: Image classification accuracy
• COCO: Object detection and segmentation
• Open Images: Large-scale object detection

• LibriSpeech: Speech recognition accuracy
• CommonVoice: Multilingual speech recognition
• AudioSet: Audio classification tasks

• Text: GPT-J 6B, BLOOM 7B
• Images: Stable Diffusion v1.5
• Music: MusicGen Large

• Customer Support: DialoGPT, RoBERTa sentiment analysis
• Document Processing: LayoutLM, TrOCR
• Search & Retrieval: CLIP, Sentence Transformers

• Multimodal: CLIP, BLIP, ALIGN
• Code: CodeT5, CodeBERT, CodeGen
• Scientific: SciBERT, BioBERT, DistilRoBERTa

We welcome contributions! Here's how you can help:

1. Add new models: Submit a PR with model details following our format
2. Update benchmarks: Share performance results and comparisons
3. Fix issues: Report broken links or outdated information
4. Improve descriptions: Make model descriptions more accurate and helpful

• Use the existing format for consistency
• Include direct Hugging Face Hub links
• Verify that demos work before submitting
• Add appropriate tags and categories
• Keep descriptions concise but informative

To be included, models should be:

• Publicly available on Hugging Face Hub
• Well-documented with clear use cases
• Actively maintained or widely adopted
• Demonstrate strong performance on benchmarks

This repository is licensed under the Creative Commons Attribution 4.0 International Public License. See LICENSE for details.

© Jehoshua 2025

⭐ Star this repository if you find it helpful!

📫 Questions or suggestions? Open an issue or start a discussion.

🔗 Share with others who might benefit from this curated list.

Last updated: August 2025