datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
fineweb-edu
📚 FineWeb-Edu
1.3 trillion tokens of the finest educational data the 🌐 web has to offer
Paper: https://arxiv.org/abs/2406.17557
What is it?
📚 FineWeb-Edu dataset consists of 1.3T tokens and 5.4T tokens (FineWeb-Edu-score-2) of educational web pages filtered from 🍷 FineWeb dataset. This is the 1.3 trillion version.
To enhance FineWeb's quality, we developed an educational quality classifier using annotations generated by LLama3-70B-Instruct. We… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu.fineweb
🍷 FineWeb
15 trillion tokens of the finest data the 🌐 web has to offer
What is it?
The 🍷 FineWeb dataset consists of more than 18.5T tokens (originally 15T tokens) of cleaned and deduplicated english web data from CommonCrawl. The data processing pipeline is optimized for LLM performance and ran on the 🏭 datatrove library, our large scale data processing library.
🍷 FineWeb was originally meant to be a fully open replication of 🦅 RefinedWeb, with a… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb.MATH-500
Dataset Card for MATH-500
This dataset contains a subset of 500 problems from the MATH benchmark that OpenAI created in their Let's Verify Step by Step paper. See their GitHub repo for the source file: https://github.com/openai/prm800k/tree/main?tab=readme-ov-file#math-splits
finephrase
Dataset Card for HuggingFaceFW/finephrase
Dataset Summary
Synthetic data generated by DataTrove:
Model: HuggingFaceTB/SmolLM2-1.7B-Instruct (main)
Source dataset: HuggingFaceFW/fineweb-edu, config sample-350BT, split train
Generation config: temperature=1.0, top_p=1.0, top_k=50, max_tokens=2048, model_max_context=8192
Speculative decoding: {"method":"suffix","num_speculative_tokens":32}
System prompt: None
Input column: text
Prompt families:
faq prompt
Rewrite… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finephrase.stack-v3-train
🥞 The Stack v3
What is it?
What is being released
How to download and use it
Dataset statistics
Dataset structure
Dataset creation
Considerations for using the data
Additional information
What is it?
The Stack v3 is the largest, most up-to-date open dataset of source code, crawled directly from GitHub and built to pre-train code LLMs with full-repository context. It is the successor to The Stack v2 and, like its predecessor, is released to make the training… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceCode/stack-v3-train.ultrachat_200k
Dataset Card for UltraChat 200k
Dataset Description
This is a heavily filtered version of the UltraChat dataset and was used to train Zephyr-7B-β, a state of the art 7b chat model.
The original datasets consists of 1.4M dialogues generated by ChatGPT and spanning a wide range of topics. To create UltraChat 200k, we applied the following logic:
Selection of a subset of data for faster supervised fine tuning.
Truecasing of the dataset, as we observed around 5% of… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k.fineweb-2
🥂 FineWeb2
A sparkling update with 1000s of languages
What is it?
This is the second iteration of the popular 🍷 FineWeb dataset, bringing high quality pretraining data to over 1000 🗣️ languages.
The 🥂 FineWeb2 dataset is fully reproducible, available under the permissive ODC-By 1.0 license and extensively validated through hundreds of ablation experiments.
In particular, on the set of 9 diverse languages we used to guide our processing decisions, 🥂… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb-2.no_robots
Dataset Card for No Robots 🙅♂️🤖
Look Ma, an instruction dataset that wasn't generated by GPTs!
Dataset Summary
No Robots is a high-quality dataset of 10,000 instructions and demonstrations created by skilled human annotators. This data can be used for supervised fine-tuning (SFT) to make language models follow instructions better. No Robots was modelled after the instruction dataset described in OpenAI's InstructGPT paper, and is comprised mostly of single-turn… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/no_robots.finepdfs
Liberating 3T of the finest tokens from PDFs
What is this?
As we run out of web pages to process, the natural question has always been: what to do next? Only a few knew about a data source that everyone avoided for ages, due to its incredible extraction cost and complexity: PDFs.
📄 FinePDFs is exactly that. It is the largest publicly available corpus sourced exclusively from PDFs, containing about 3 trillion tokens across 475 million documents in 1733 languages.
Compared to… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finepdfs.finetranslations
💬 FineTranslations
The world's knowledge in 1+1T tokens of parallel text
What is it?
This dataset contains over 1 trillion tokens of parallel text in English and 500+ languages. It was obtained by translating data from 🥂 FineWeb2 into English using Gemma3 27B.
We relied on datatrove's inference runner to deploy a synthetic data pipeline at scale. Its checkpointing and VLLM lifecycle management features allowed us to use leftover compute from the HF cluster… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finetranslations.ultrafeedback_binarized
Dataset Card for UltraFeedback Binarized
Dataset Description
This is a pre-processed version of the UltraFeedback dataset and was used to train Zephyr-7Β-β, a state of the art chat model at the 7B parameter scale.
The original UltraFeedback dataset consists of 64k prompts, where each prompt is accompanied with four model completions from a wide variety of open and proprietary models. GPT-4 is then used to assign a score to each completion, along criteria like… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized.CodeAlpaca_20KThis dataset splits the original CodeAlpaca dataset into train and test splits.
fineweb-edu-score-2
📚 FineWeb-Edu-score-2
1.3 trillion tokens of the finest educational data the 🌐 web has to offer
What is it?
📚 FineWeb-Edu dataset consists of 1.3T tokens (FineWeb-Edu) and 5.4T tokens of educational web pages filtered from 🍷 FineWeb dataset. This is the 5.4 trillion version.
Note: this version uses a lower educational score threshold = 2, which results in more documents, but lower quality compared to the 1.3T version. For more details check the… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu-score-2.finepdfs-edu
📚 FinePDFs-Edu
350B+ of highly educational tokens from PDFs 📄
What is it?
📚 FinePDFs-Edu dataset consists of 350B+ tokens of educational PDFs filtered from 📄 FinePDFs dataset covering 69 languages.
FinePDFs was created using the formula inspired from FineWeb-Edu, we developed an educational quality classifier using annotations generated by Qwen3-235B-A22B-Instruct-2507 for each of 69 languages present in this dataset.
We then used this classifier to retain only the… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finepdfs-edu.REDDIT_comments
Dataset Card for "REDDIT_comments"
Dataset Summary
Comments of 50 high-quality subreddits, extracted from the REDDIT PushShift data dumps (from 2006 to Jan 2023).
Supported Tasks
These comments can be used for text generation and language modeling, as well as dialogue modeling.
Dataset Structure
Data Splits
Each split corresponds to a specific subreddit in the following list: "tifu", "explainlikeimfive", "WritingPrompts", "changemyview"… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceGECLM/REDDIT_comments.finewiki
This is an updated and better extracted version of the wikimedia/Wikipedia dataset originally released in 2023. We carefully parsed Wikipedia HTML dumps from August of 2025 covering 325 languages.
This dataset:
fully renders templates as it was extracted from HTML and not markdown dumps
removes redirects, disambiguation, and other non main article pages
includes detailed metadata such as page ID, title, last modified date, wikidate ID, version and markdown version of the textpreserves… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finewiki.Multilingual-Thinking
Dataset summary
Multilingual-Thinking is a reasoning dataset where the chain-of-thought has been translated from English into one of 4 languages: Spanish, French, Italian, and German. The dataset was created by sampling 1k training samples from the SystemChat subset of SmolTalk2 and translating the reasoning traces with another language model.
This dataset was used in the OpenAI Cookbook to fine-tune the OpenAI gpt-oss models.
You can load the dataset using:
from datasets import… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/Multilingual-Thinking.REDDIT_submissions
Dataset Card for "REDDIT_submissions"
Dataset Summary
Submissions of 50 high-quality subreddits, extracted from the REDDIT PushShift data dumps (from 2006 to Jan 2023).
Supported Tasks
These submissions can be used for text generation and language modeling, as well as dialogue modeling.
Dataset Structure
Data Splits
Each split corresponds to a specific subreddit in the following list: "tifu", "explainlikeimfive", "WritingPrompts"… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceGECLM/REDDIT_submissions.CXM_Arena
Dataset Card for CXM Arena Benchmark Suite
Dataset Description
This dataset, "CXM Arena Benchmark Suite," is a comprehensive collection designed to evaluate various AI capabilities within the Customer Experience Management (CXM) domain. It consolidates five distinct tasks into a unified benchmark, enabling robust testing of models and pipelines in business contexts. The entire suite was synthetically generated using advanced large language models, primarily… See the full description on the dataset page: https://huggingface.co/datasets/sprinklr-huggingface/CXM_Arena.finetranslations-edu
💬 FineTranslations
The world's knowledge in 1+1T tokens of parallel text
NOTE: this is the Edu version of the dataset, containing only the top 10% scoring data based on an educational classifier applied to the English translations. It has no splits. For the base dataset, see here.
What is it?
This dataset contains over 1 trillion tokens of parallel text in English and 500+ languages. It was obtained by translating data from 🥂 FineWeb2 into English using… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finetranslations-edu.carbon-pretraining-corpus
🧬 Carbon Pretraining Corpus
Description
173M DNA & RNA sequences · 1.1 trillion nucleotides — the DNA pretraining mixture used to train Carbon, a genomic foundation model.
This dataset is a collection of data sources intended for training genomic foundation models, such as Carbon. It contains DNA and RNA sequences spanning eukaryote and prokaryote species.
Across the four main configs it totals 1.1 T DNA base pairs (180B tokens with Carbon's 6-mer tokenizer). A… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceBio/carbon-pretraining-corpus.orca_dpo_pairs
Dataset Card for Orca DPO Pair
Dataset Description
This is a pre-processed version of the OpenOrca dataset.
The original OpenOrca dataset is a collection of augmented FLAN data that aligns, as best as possible, with the distributions outlined in the Orca paper.
It has been instrumental in generating high-performing preference-tuned model checkpoints and serves as a valuable resource for all NLP researchers and developers!
Dataset Summary
The OrcaDPO Pair… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/orca_dpo_pairs.tau2-bench-data
Dataset summary
This dataset contains the domain data from the tau2-bench repository for agentic evaluation.
To download it in your code, run:
hf download HuggingFaceH4/tau2-bench-data --repo-type dataset --local-dir data/tau2
HuggingFaceTB-cosmopedia-cnhelpful_instructionsHelpful Instructions is a dataset of (prompt, completion) pairs that are derived from a variety of public datasets. As the name suggests, it focuses on instructions that are "helpful", i.e. the kind of questions or tasks a human user might instruct an AI assistant to perform.huggingface_myanmar_english_translation
Cleaned & Sorted Myanmar-English Translation Dataset
This dataset is a cleaned, Unicode-normalized, and sorted version of the Myanmar (Burmese) subset from the massive FineTranslations dataset.
While the original dataset is excellent, Myanmar text on the web is often a mix of standard Unicode and the non-standard Zawgyi encoding. This repository fixes those encoding issues to provide a high-quality dataset for NLP tasks.
Key Improvements in this Version
Zawgyi… See the full description on the dataset page: https://huggingface.co/datasets/freococo/huggingface_myanmar_english_translation.HuggingFaceFW-finewiki-sample
HuggingFaceFW/finewiki sample
A uniformly randomized subset of HuggingFaceFW/finewiki, created to provide a smaller and more manageable dataset for analysis, fine-tuning, and benchmarking.
Overview
This sample includes Wikipedia articles from languages with more than one million pages. Sampling is performed uniformly at random instead of alphabetically to ensure unbiased representation.
Language Inclusion Criteria
Languages were selected based on page count and… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/HuggingFaceFW-finewiki-sample.self_instructThis dataset splits the original Self-instruct dataset into training (90%) and test (10%).
tokbench-corpora
tokbench corpora
The input corpora for tokbench, a
benchmark that measures tokenizer implementations against each other on the
same bytes. Each config is one corpus: ~5 MB of real text chosen to stress a
different part of a tokenizer.
Nothing here is new text. It is a fixed, pinned, redistributable excerpt of
public datasets, packaged so a tokenizer benchmark is reproducible by anyone
without re-deriving the inputs. Provenance and licence for every config are in
the table below.… See the full description on the dataset page: https://huggingface.co/datasets/huggingface/tokbench-corpora.OpenHermes-2.5-1k-longest
Dataset Card for OpenHermes-2.5-1k-longest
OpenHermes-2.5-1k-longest is a dataset of 1,000 samples derived from teknium/OpenHermes-2.5 using the Long is More for Alignment protocol. This protocol consists of selecting the 1,000 longest responses and provides a strong baseline to measure performance against. For example, fine-tuning mistralai/Mistral-7B-v0.1 on this dataset using similar hyperparameters to those given in the paper produces a chat model that achieves a score of ~7.0… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceH4/OpenHermes-2.5-1k-longest.