Skip to content

Datasets

Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,797 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

DatasetGD-ML/TransitLM

Public transit route planning

Data for training language models in public transit route planning across Chinese cities.

  • Text generation
  • Chinese
919 84100K–1M rows
Datasetllm-jp/magpie-sft-v1.0

Magpie Japanese Synthetic SFT Dataset

100k to 1M Japanese instruction-response pairs for SFT generated by Calm3 and Qwen2.5 models.

  • Text generation
  • Japanese
915 23100K–1M rows
Datasetdaekeun-ml/naver-news-summarization-ko

Korean news with summaries

27,400 news text and summary pairs in Korean for summarization model training and testing.

  • Summarization
  • Korean
909 7010K–100K rows
Datasetnvidia/Nemotron-Personas-Brazil

Synthetic personas of Brazil

A large Portuguese-language dataset of synthetic personas aligned with real Brazilian distributions.

  • Text generation
  • Portuguese
907 1071M–10M rows
DatasetOpenDataArena/MMFineReason-SFT-123K-Qwen3-VL-235B-Thinking

MMFineReason: complex VQA examples for SFT

123k most difficult image-text examples from MMFineReason where Qwen3-VL-4B-Thinking consistently fails.

  • Visual question answering
  • Question answering
  • Text generation
  • English
907 87100K–1M rows
DatasetDSFFGFG456/fable-5-coding-and-debugging-traces

Claude Fable 5 agentic coding trajectories

The dataset contains instruction, tool use, and agentic coding trajectories from Claude Fable 5.

  • Text generation
  • English
902 810K–100K rows
Datasetfacebook/ExploreToM

Meta benchmark for testing theory of mind in LLMs

Synthetic dataset from Facebook for evaluating and training LLM ability to understand mental states (Theory of Mind).

  • Question answering
  • English
899 4910K–100K rows
DatasetModotte/CodeX-2M-Thinking

CodeX-2M-Thinking Synthetic SFT Dataset

Synthetic English texts for instruction tuning and fine-tuning response generation models.

  • Text generation
  • Question answering
  • English
895 1311M–10M rows
DatasetSunbird/salt

African multilingual speech and text corpus

Parallel corpus of texts and audio recordings in English and six East African languages.

  • Automatic speech recognition
  • Text → speech
  • Translation
  • Acoli
  • English
  • lgg
  • +6
892 2710K–100K rows🔒 on request
DatasetDen4ikAI/russian_dialogues_2

Russian dialogues for chatbot training

A corpus of 1.6 million contextual Russian dialogues for training conversational agents.

  • Text generation
  • Russian
884 201M–10M rows
DatasetRyokoAI/Fandom23K

Articles from 23 thousand Fandom wiki projects

A massive English corpus of over 15.6 million articles from Fandom.com for pre-training and text generation.

  • Text classification
  • Text generation
  • English
882 321M–10M rows
Datasetstepfun-ai/PaCoRe-Train-8k

PaCoRe data for parallel reasoning

About 8K text examples for training models in parallel coordinated reasoning within PaCoRe.

  • Text generation
  • English
881 811K–10K rows
DatasetReasonCore/open-spatial-reasoning

Road camera 3D spatial reasoning benchmark

Multiple-choice visual question dataset for testing 3D spatial understanding in multimodal models.

  • Multiple choice
  • English
875 66up to 1K rows
Datasetronantakizawa/github-codereview

Code reviews and edits from top GitHub repositories

A large collection of pull request reviewer comments and subsequent code fixes for training automated code review models.

  • Text generation
  • English
  • Code
874 66100K–1M rows
Datasetangeluriot/french_instruct

French instructions and dialogues

276k instruction-response pairs and multi-turn dialogues in French for language model training.

  • Question answering
  • Text generation
  • Text classification
  • French
868 21100K–1M rows
Datasetlaolao77/MMDU

Multi-image multi-turn dialogue benchmark

A set of 110 complex dialogues and over 1600 questions across image series for evaluating multimodal models.

  • Visual question answering
  • Question answering
  • English
863 44up to 1K rows
Datasetianncity/KIMI-K2.5-1000000x

1M reasoning traces from KIMI-K2.5

One million reasoning chains distilled from KIMI-K2.5, broken down by topic.

  • Text generation
  • Question answering
  • English
857 275100K–1M rows
DatasetAlibaba-Apsara/Superior-Reasoning-SFT-gpt-oss-120b-Logprob

Reasoning log-probabilities from gpt-oss-120b model

Token-level log-probabilities from the gpt-oss-120b teacher model for distilling English reasoning skills.

  • Text generation
  • English
855 63100K–1M rows
DatasetJackrong/Kimi-K2.5-Reasoning-1M-Cleaned

Cleaned reasoning dataset from Kimi-K2.5 model

Cleaned English and Chinese SFT sample with reasoning chains from KIMI-K2.5.

  • Text generation
  • Question answering
  • English
  • Chinese
853 36100K–1M rows
Datasetneulab/MultiUITop pick

MultiUI multimodal web interface dataset (7.3M)

7.3 million multimodal examples from 1 million websites for training web agents and VLM models.

  • Question answering
  • Visual question answering
  • English
847 531M–10M rows🔒 on request