The pipeline() function is the easiest way to run pre-trained models — it handles tokenization, inference, and post-processing automatically.

Available Pipeline Tasks

from transformers import pipeline

# Text tasks
pipeline("text-generation")       # Generate text
pipeline("summarization")         # Summarize documents  
pipeline("translation_en_to_fr")  # Translate
pipeline("sentiment-analysis")    # Classify sentiment
pipeline("zero-shot-classification") # Classify without training
pipeline("question-answering")    # Q&A from context
pipeline("ner")                   # Named entity recognition

# Multimodal
pipeline("image-classification")  # Classify images
pipeline("automatic-speech-recognition") # Transcribe audio

Reference:

Pipeline tutorial

https://huggingface.co/docs/transformers/pipeline_tutorial