Skip to content

Datasets

Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.

Data type

For which models

Find a dataset for your goal

Answered by Gemini 4 · the chat is kept only in this tab

Datasetmath-ai/AutoMathText-2.5Top pick

AutoMathText-2.5 math and STEM corpus with 2T tokens

Cleaned bilingual (EN/ZH) corpus exceeding 2 trillion tokens on mathematics, code, and reasoning for LLM training.

  • Text generation
  • Question answering
  • English
  • Chinese
39.4K 31B–10B rows
Datasetvoilaj/swiss-caselaw

Swiss Case Law in Three Languages (1M+ Decisions)

The largest open corpus of over 1 million Swiss court decisions for legal NLP tasks.

  • Text classification
  • Summarization
  • Question answering
  • German
  • French
  • Italian
37.4K 151M–10M rows
Datasetallenai/WildChat-1MTop pick

One million human-ChatGPT dialogues

Real multilingual user conversations with ChatGPT featuring demographics and prompt metadata.

  • Text generation
  • Question answering
37.4K 486100K–1M rows
DatasetPleIAs/SYNTHTop pick

Synthetic corpus for reasoning models

79.6 million Wikipedia-based text examples for end-to-end training of small reasoning models.

  • Text generation
  • Zero shot classification
  • Summarization
  • English
  • French
  • Italian
  • +5
37.2K 31210M–100M rows
Datasetlianghsun/tw-legal-oa-corpus

Taiwan legal corpus in Traditional Chinese

Over 33 thousand open legal articles and publications in Traditional Chinese for fine-tuning language models.

  • Text generation
  • Chinese
35.8K 010K–100K rows🔒 on request
Datasetmicrosoft/orca-math-word-problems-200k

School math word problems with answers

About 200k English school math word problems with answers generated by GPT-4-Turbo.

  • Question answering
  • English
35.4K 520100K–1M rows
Datasetsedthh/gutenberg_english

English books from Project Gutenberg

Collection of 48,284 English books from Project Gutenberg, prepared for generative model training.

  • Text generation
  • English
35.4K 4110K–100K rows
Datasetdatabricks/officeqa

Databricks benchmark for complex financial documents

Q&A based on historical US Treasury bulletins to evaluate AI agents' ability to analyze tables, charts, and text.

  • Question answering
  • Text generation
  • Text retrieval
  • English
35.2K 42up to 1K rows🔒 on request
Datasetespnet/yodas-granaryTop pick

YODAS-Granary European speech corpus for ASR and translation

A large multilingual audio dataset (10M-100M entries) with pseudo-labels for European speech recognition and translation.

  • Automatic speech recognition
  • Translation
  • Bulgarian
  • Czech
  • Danish
  • +9
35.1K 4310M–100M rows
Datasetnvidia/Nemotron-CC-v2.1Top pick

Nemotron-CC v2.1 pretraining corpus

Updated data for LLM pretraining: math, code, English Common Crawl, and synthetic data.

  • Text generation
35K 1471B–10B rows🔒 on request
DatasetHuggingFaceH4/ultrafeedback_binarized

Preference pairs for chat models

Processed UltraFeedback: 64k prompts with model answers and GPT-4 scores aggregated into preference pairs.

  • Text generation
  • English
35K 363100K–1M rows
Datasetlukakuma/web-trajectories-preference-raw

Raw web agent trajectories with images

Large raw archive of successful and failed web agent actions in English and Korean for subsequent filtering.

  • Text generation
  • English
  • Korean
34.4K 01M–10M rows🔒 on request
Datasetgoogle-research-datasets/nq_open

NQ-Open open-domain question answering benchmark

Classic English benchmark for testing model fact-answering capability without source context.

  • Question answering
  • English
34.4K 3810K–100K rows
Datasetproject-numina/NuminaMath-1.5

NuminaMath 1.5: Math with CoT

Around 900k competitive math problems with Chain of Thought solutions.

  • Text generation
  • English
34.1K 210100K–1M rows
Datasetopenbmb/UltraData-MathTop pick

290B-token math pre-training corpus

290 billion tokens of English and Chinese math data across three tiers for LLM pre-training.

  • Text generation
  • English
  • Chinese
34K 365100M–1B rows
DatasetTevatron/browsecomp-plus

Deep-Research benchmark for deep search systems

Test set of complex search queries and a fixed corpus of 100 thousand web documents for fair agent comparison.

  • Question answering
33.7K 41up to 1K rows
Datasetlongvideobench/LongVideoBench

Long-video understanding benchmark up to 1 hour for LMMs

3,763 long captioned videos with questions to evaluate long-context multimodal models.

  • Multiple choice
  • Visual question answering
  • English
33.6K 641K–10K rows🔒 on request
DatasetPleIAs/US-PD-NewspapersTop pick

US Public Domain Newspapers Archive

A massive corpus of digitized historical US newspapers in English for training language models.

  • Text generation
  • English
33.5K 5010M–100M rows
Datasetwulin222/MME-Unify

MME-Unify benchmark for multimodal models

Evaluation set of images and English text across 10 tasks and 30 subtasks for testing U-MLLMs.

  • Multiple choice
  • Question answering
  • Visual question answering
  • English
32.7K 31K–10K rows
DatasetTree-AI-lab/smollm-corpus-multilingueTop pick

Multilingual corpus with rare African languages

A text corpus of 10 to 100M records for pre-training models in English, French, and West African languages.

  • Text generation
  • English
  • French
  • bci
  • +2
32.4K 210M–100M rows🔒 on request