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

DatasetSciPhi/AgentSearch-V1Top pick

Corpus and embeddings for agent search

Over 50 million documents and 1 billion passages with jina-v2-base embeddings for training search.

  • Text generation
  • English
3K 9210K–100K rows
Datasetnyuuzyou/google-code-archiveTop pick

Google Code Archive Open Source Snapshot

Google Code project snapshot from 2006-2016 for training models on historical code.

  • Text generation
  • Code
  • English
3K 7410M–100M rows
Dataseta-m-team/AM-DeepSeek-Distilled-40MTop pick

Large distilled reasoning corpus for LLMs

A large-scale text dataset with difficulty grading for SFT and enhancing model reasoning skills.

  • Text generation
  • Chinese
  • English
3K 5610M–100M rows
DatasetBAAI/DenseFusion-1MTop pick

1M Dataset with Detailed Image Descriptions

A large multimodal set of 1 million to 10 million image-text pairs in English.

  • Question answering
  • Visual question answering
  • English
3K 441M–10M rows
DatasetOpenMed/Medical-Reasoning-SFT-MegaTop pick

Medical SFT with chain-of-thought

1.79M medical examples with reasoning from 7 models for SFT.

  • Text generation
  • Question answering
  • English
3K 1011M–10M rows
Datasetajibawa-2023/JavaScript-Code-Large

Large JavaScript Source Code Corpus

About 5 million JavaScript code files for pre-training and fine-tuning code models.

  • Text generation
  • English
3K 341M–10M rows
Datasetmirobody/MetAgentBench

Metabolic Medical Agent Benchmark

Interactive clinical benchmark for evaluating AI agents in metabolic medicine and GLP-1 therapy.

  • Question answering
  • English
  • Chinese
3K 6up to 1K rows
Datasetnvidia/Nemotron-Pretraining-Code-v1

Math and code for LLM pre-training

Pre-training corpus for generative models featuring math, code, and multilingual Q&A.

  • Text generation
3K 87100M–1B rows🔒 on request
Datasetkaist-ai/CoT-Collection

Chain of Thought Collection

1 million to 10 million English examples with chains of thought for generation and classification.

  • Text generation
  • Text classification
  • English
3K 1731M–10M rows
DatasetFreedomIntelligence/CMB

Chinese medical benchmark

Comprehensive test set of medical questions and answers for evaluating language models in Chinese.

  • Question answering
  • Text generation
  • Chinese
3K 38up to 1K rows
DatasetAlicanKiraz0/All-CVE-Records-Training-Dataset

CVE vulnerability multi-turn dataset (1999–2025)

Around 300 thousand CVE vulnerability records formatted as multi-turn dialogues for AI agent training.

  • Text generation
  • English
2.9K 63100K–1M rows
Datasethfl/cmrc2018

Chinese Wikipedia Question Answering

About 20,000 question-answer pairs with span-based answers for Chinese reading comprehension.

  • Question answering
  • Chinese
2.9K 3910K–100K rows
Datasetopen-index/open-arxiv

Complete metadata of nearly 3 million arXiv papers

Metadata of 2.99 million scientific publications from 1991 to 2026 in parquet format for search and analytics.

  • Text classification
  • Text generation
  • Embeddings
  • English
  • Multilingual
2.9K 211M–10M rows
Datasetgbharti/finance-alpaca

Financial Alpaca instructions

Finance instruction-response pairs based on Alpaca and FiQA plus 1.3k synthetic examples.

  • Text generation
  • English
2.9K 16610K–100K rows
DatasetAlicanKiraz0/Cybersecurity-Dataset-Fenrir-v2.1

Cybersecurity instructions for SFT

99,870 system-user-assistant triplets for defensive cybersecurity SFT in English.

  • Text generation
  • English
2.9K 17310K–100K rows
DatasetHuggingFaceBio/carbon-pretraining-corpusTop pick

1.1 trillion nucleotide DNA and RNA genomic corpus

173 million eukaryotic and prokaryotic DNA and RNA sequences for pretraining base genomic models.

  • Text generation
2.9K 34100M–1B rows
DatasetCohereLabs/aya_evaluation_suite

Multilingual dialogue prompt benchmark by Cohere

26,750 open prompts in over 100 languages for evaluating text generation quality.

  • Text generation
  • Afrikaans
  • Albanian
  • Amharic
  • +9
2.8K 5510K–100K rows
Datasetbigcode/the-stack-v2-dedup

Deduplicated code corpus The Stack v2

Nearly deduplicated version of The Stack v2—a large code corpus for LLM training, grouped by repositories.

  • Text generation
  • Code
2.8K 1481B–10B rows🔒 on request
Datasetnvidia/Nemotron-Personas-GermanyTop pick

Synthetic personas of German residents by NVIDIA

1 to 10 million descriptions of virtual German residents generated based on real demographics.

  • Text generation
  • German
2.8K 531M–10M rows
DatasetHPAI-BSC/medical-specialities

Medical question classification

Dataset for evaluating medical language models across 35 categories.

  • Question answering
  • English
2.7K 810K–100K rows
Datasetshibing624/alpaca-zh

50k Chinese Alpaca instructions

About 50k instruction-response pairs in Chinese generated by GPT-4 using the Alpaca method.

  • Text generation
  • Chinese
2.7K 14610K–100K rows
DatasetRodrigoFlorencio/duvidas-saude-ptbr

Portuguese health Q&A corpus

Over 33,000 real patient questions and 127,000 doctor answers in Portuguese.

  • Question answering
  • Text generation
  • Portuguese
2.7K 2100K–1M rows