caoyuji/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, 🥂 FineWeb2 outperforms other popular pretraining datasets covering multiple languages (such as CC-100, mC4, CulturaX or HPLT, while being substantially larger) and, in some cases, even performs better than some datasets specifically curated for a single one of these languages, in our diverse set of carefully selected evaluation tasks: FineTasks.

The data was sourced from 96 CommonCrawl snapshots, spanning the summer of 2013 to April 2024, and processed using 🏭 datatrove, our large scale data processing library. This carefully deduplicated and filtered dataset comprises roughly 20 terabytes, across 5 billion documents, with over 3 trillion words (see How many tokens? for more details). For PII and opt-out see Personal and Sensitive Information and opt-out.

You will find our ablation and evaluation setup in this github repo. We will soon upload model checkpoints from our ablation experiments.

Read our 📝 research paper for details on the dataset creation!

Languages and available subsets

For English data, please refer to the original 🍷 FineWeb.

Each language is identified by its ISO 639-3 code, and the data is grouped by language-script pairs, since some languages have content in multiple scripts.

In total, we provide filtered data for 1,868 language-script pairs. Of these, 474 have more than 1 thousand documents, and 203 have more than 10 thousand documents of filtered data. Most languages also include a small test split which should not be trained on.

While we tried our best to not overfilter, we know that our filtering isn't perfect, and wanted to allow the community to easily re-filter the data with their own filtering criteria. We have therefore also uploaded the data that was removed by our filtering pipeline for each language (it is suffixed by _removed). The filtered + the removed subsets of each language represent the entire data for a given language following global deduplication, which means that you do not have to re-deduplicate it yourself. You can find and adapt our filtering code here. The removed data is available through direct download (using hub_download for example) but not through load_dataset, as there would otherwise be an excessive number of subsets.

Additionally, we also uploaded data for scripts that the language classifier does not support or in a supported script but unknown language, without any deduplication or filtering. These are prefixed by und_.

The following table shows the size of the filtering subset for the biggest 80 languages. The full list is available on Github.

ISO 639-3 code	Script	Name	Language Family	Subset	Words	Documents	UTF-8 Bytes	Disk size
rus	Cyrl	Russian	Indo-European	rus_Cyrl	588,579,493,780	699,083,579	5.82TB	1.81TB
cmn	Hani	Mandarin Chinese	Sino-Tibetan	cmn_Hani	543,543,038,750	636,058,984	2.42TB	1.48TB
deu	Latn	German	Indo-European	deu_Latn	262,271,052,199	495,964,485	1.51TB	719.08GB
jpn	Jpan	Japanese	Japonic	jpn_Jpan	331,144,301,801	400,138,563	1.50TB	667.44GB
spa	Latn	Spanish	Indo-European	spa_Latn	261,523,749,595	441,287,261	1.32TB	593.82GB
fra	Latn	French	Indo-European	fra_Latn	220,662,584,640	360,058,973	1.11TB	502.82GB
ita	Latn	Italian	Indo-European	ita_Latn	139,116,026,491	238,984,437	739.24GB	332.47GB
por	Latn	Portuguese	Indo-European	por_Latn	109,536,087,117	199,737,979	569.24GB	256.92GB
pol	Latn	Polish	Indo-European	pol_Latn	73,119,437,217	151,966,724	432.01GB	210.35GB
nld	Latn	Dutch	Indo-European	nld_Latn	74,634,633,118	147,301,270	397.51GB	176.98GB
ind	Latn	Indonesian	Austronesian	ind_Latn	60,264,322,142	100,238,529	348.65GB	141.70GB
vie	Latn	Vietnamese	Austro-Asiatic	vie_Latn	50,886,874,358	61,064,248	319.83GB	121.19GB
fas	Arab	Persian	Indo-European	fas_Arab	39,705,799,658	58,843,652	304.62GB	95.33GB
arb	Arab	Standard Arabic	Afro-Asiatic	arb_Arab	32,812,858,120	61,977,525	293.59GB	98.69GB
tur	Latn	Turkish	Turkic	tur_Latn	41,933,799,420	95,129,129	284.52GB	125.53GB
tha	Thai	Thai	Kra-Dai	tha_Thai	24,662,748,945	35,897,202	278.68GB	69.91GB
ukr	Cyrl	Ukrainian	Indo-European	ukr_Cyrl	25,586,457,655	53,101,726	254.86GB	84.98GB
ell	Grek	Modern Greek (1453-)	Indo-European	ell_Grek	22,827,957,288	47,421,073	222.05GB	73.16GB
kor	Hang	Korean	Koreanic	kor_Hang	48,613,120,582	60,874,355	213.43GB	98.50GB
ces	Latn	Czech	Indo-European	ces_Latn	35,479,428,809	66,067,904	206.33GB	102.38GB
swe	Latn	Swedish	Indo-European	swe_Latn	35,745,969,364	59,485,306	202.96GB	88.63GB
hun	Latn	Hungarian	Uralic	hun_Latn	30,919,839,164	49,935,986	199.69GB	91.73GB
ron	Latn	Romanian	Indo-European	ron_Latn	35,017,893,659	58,303,671	186.19GB	85.37GB
nob	Latn	Norwegian Bokmål	Indo-European	nob_Latn	32,008,904,934	38,144,343	172.05GB	78.25GB
dan	Latn	Danish	Indo-European	dan_Latn	28,055,948,840	45,391,655	150.72GB	65.74GB
bul	Cyrl	Bulgarian	Indo-European	bul_Cyrl	16,074,326,712	25,994,731	145.75GB	45.68GB
fin	Latn	Finnish	Uralic	fin_Latn	20,343,096,672	36,710,816	143.03GB	61.94GB
hin	Deva	Hindi	Indo-European	hin_Deva	11,173,681,651	22,095,985	120.98GB	31.92GB
ben	Beng	Bengali	Indo-European	ben_Beng	6,153,579,265	15,185,742	87.04GB	22.25GB
slk	Latn	Slovak	Indo-European	slk_Latn	14,808,010,769	29,991,521	85.43GB	43.00GB
heb	Hebr	Hebrew	Afro-Asiatic	heb_Hebr	8,462,976,117	14,491,748	68.71GB	23.15GB
lit	Latn	Lithuanian	Indo-European	lit_Latn	9,132,828,961	13,471,965	56.50GB	25.75GB
bos	Latn	Bosnian	Indo-European	bos_Latn	9,086,837,979	21,243,255	49.18GB	24.61GB
slv	Latn	Slovenian	Indo-European	slv_Latn	7,688,373,264	12,059,130	41.80GB	19.22GB
ekk	Latn	Standard Estonian	Uralic	ekk_Latn	6,564,292,000	10,218,587	40.82GB	18.35GB
cat	Latn	Catalan	Indo-European	cat_Latn	8,348,091,726	17,136,414	40.35GB	18.52GB
tam	Taml	Tamil	Dravidian	tam_Taml	1,937,150,898	5,528,854	36.97GB	8.79GB
hrv	Latn	Croatian	Indo-European	hrv_Latn	6,609,299,440	6,195,824	35.91GB	16.36GB
lvs	Latn	Standard Latvian	Indo-European	lvs_Latn	5,371,151,279	8,030,316	33.36GB	14.70GB
zsm	Latn	Standard Malay	Austronesian	zsm_Latn	5,648,387,840	9,421,248	31.94GB	13.28GB
azj	Latn	North Azerbaijani	Turkic	azj_Latn	3,894,255,826	7,291,231	26.90GB	10.49GB
srp	Cyrl	Serbian	Indo-European	srp_Cyrl	2,858,500,314	4,146,124	26.87GB	8.64GB
kat	Geor	Georgian	Kartvelian	kat_Geor	1,439,572,993	3,706,659	25.23GB	6.33GB
npi	Deva	Nepali (individual language)	Indo-European	npi_Deva	1,642,856,349	4,888,163	25.13GB	6.22GB
mar	Deva	Marathi	Indo-European	mar_Deva	1,541,225,070	3,912,702	22.57GB	5.85GB
mal	Mlym	Malayalam	Dravidian	mal_Mlym	1,054,187,581	3,322,526	22.27GB	5.51GB
kaz	Cyrl	Kazakh	Turkic	kaz_Cyrl	1,876,843,453	3,344,366	20.67GB	6.33GB
urd	Arab	Urdu	Indo-European	urd_Arab	2,733,266,493	4,809,542	19.93GB	6.40GB
als	Latn	Tosk Albanian	Indo-European	als_Latn	3,454,387,059	8,597,826	18.18GB	8.42GB
mkd	Cyrl	Macedonian	Indo-European	mkd_Cyrl	1,611,392,841	4,150,902	14.99GB	4.82GB
tel	Telu	Telugu	Dravidian	tel_Telu	891,002,487	1,964,395	14.42GB	3.68GB
kan	Knda	Kannada	Dravidian	kan_Knda	748,850,327	2,390,982	12.91GB	3.28GB
mya	Mymr	Burmese	Sino-Tibetan	mya_Mymr	854,400,671	1,558,304	12.35GB	2.90GB
guj	Gujr	Gujarati	Indo-European	guj_Gujr	934,124,052	2,127,094	11.71GB	3.11GB
bel	Cyrl	Belarusian	Indo-European	bel_Cyrl	1,166,541,148	2,100,873	11.47GB	3.87GB
isl	Latn	Icelandic	Indo-European	isl_Latn	1,696,354,360	3,014,429	10.27GB	4.59GB
khm	Khmr	Khmer	Austro-Asiatic	khm_Khmr	667,495,692	1,586,460	8.70GB	2.12GB
khk	Cyrl	Halh Mongolian	Mongolic	khk_Cyrl	824,211,882	1,622,882	8.52GB	2.58GB
fil	Latn	Filipino	Austronesian	fil_Latn	1,636,238,017	2,349,050	8.13GB	3.34GB
ary	Arab	Moroccan Arabic	Afro-Asiatic	ary_Arab	843,523,994	2,365,405	7.74GB	2.67GB
afr	Latn	Afrikaans	Indo-European	afr_Latn	1,598,352,868	1,992,040	7.69GB	3.40GB
hye	Armn	Armenian	Indo-European	hye_Armn	634,273,060	1,757,415	7.17GB	2.26GB
sin	Sinh	Sinhala	Indo-European	sin_Sinh	512,453,069	1,185,323	7.05GB	1.87GB
glg	Latn	Galician	Indo-European	glg_Latn	1,236,233,473	2,522,814	6.47GB	2.92GB
uzn	Cyrl	Northern Uzbek	Turkic	uzn_Cyrl	544,866,919	1,357,811	6.12GB	1.83GB
pan	Guru	Panjabi	Indo-European	pan_Guru	522,788,467	944,160	5.64GB	1.47GB
ory	Orya	Odia	Indo-European	ory_Orya	333,760,951	1,298,188	4.92GB	1.28GB
uzn	Latn	Northern Uzbek	Turkic	uzn_Latn	687,002,994	1,233,463	4.45GB	1.90GB
kir	Cyrl	Kirghiz	Turkic	kir_Cyrl	397,449,282	1,069,582	4.36GB	1.37GB
eus	Latn	Basque	Language isolate	eus_Latn	711,939,889	1,569,434	4.30GB	1.90GB
lat	Latn	Latin	Indo-European	lat_Latn	714,764,848	1,473,541	3.86GB	1.64GB
tgk	Cyrl	Tajik	Indo-European	tgk_Cyrl	396,209,383	688,384	3.75GB	1.15GB
gmh	Latn	Middle High German (ca. 1050-1500)	Indo-European	gmh_Latn	506,396,917	84,495	3.41GB	1.28GB
swh	Latn	Swahili (individual language)	Niger-Congo	swh_Latn	569,542,024	1,206,300	3.08GB	1.33GB
arz	Arab	Egyptian Arabic	Afro-Asiatic	arz_Arab	345,040,810	853,290	2.92GB	1.06GB
nno	Latn	Norwegian Nynorsk	Indo-European	nno_Latn	522,740,774	1,214,870	2.68GB	1.30GB
cym	Latn	Welsh	Indo-European	cym_Latn	523,226,616	831,878	2.50GB	1.10GB
amh	Ethi	Amharic	Afro-Asiatic	amh_Ethi	239,936,286	428,373	2.49GB	848.50MB
pbt	Arab	Southern Pashto	Indo-European	pbt_Arab	337,138,269	639,983	2.41GB	816.03MB
ckb	Arab	Central Kurdish	Indo-European	ckb_Arab	236,342,609	554,993	2.39GB	783.85MB
...	...	...	...	...	...	...	...
Total					3,339,271,691,958	5,018,505,566	20.78TB	8.58TB

How many tokens?

The number of tokens obtained when tokenizing data in a specific language heavily depends on whether the tokenizer was trained with that language, and its script, in mind. For instance, while employing the gpt2 tokenizer to tokenize Thai data might result in a very large number of tokens, using a tokenizer explicitly trained for south-east asian languages would considerably bring down this number.

As such, we chose to only report total number of documents, disk size and words for each language, as reported by the word tokenizer (we don't mean gpt2 here, but a tool that will only split words) that we assigned to each language.

Changelog

Previous versions remain available in the branch version name. You can access them using for example revision="v2.0.0".

• v2.1.1 (27-10-2025): Added han_Latn and nan_Latn. Fixed features issue when using load_dataset for some languages.
• v2.1.0 (27-06-2025): Filtering was slightly changed to match the version from our paper. The dataset size has increased. We have also added additional filtering to lower-resource languages to increase precision.
• v2.0.1 (08-01-2025): We reran the "fixes" step with most fixes from FTFY disabled except encoding correction. These fixes were, for example, changing all full-width punctuation in Chinese to half-width (which is not commonly used), as well as applying other normalizations that could make models not recognize certain types of characters or formatting. See here.
• v2.0.0 (08-12-2024): Initial version

How to download and use 🥂 FineWeb2

See the tables above for the subset of the language and version (filtered or removed) of the data you want to download.

We currently do not provide smaller sample versions, but by setting limit or using streaming=True you can easily fetch a sample of the data. If there is interest from the community we might upload smaller sampled versions later on.

Using 🏭 datatrove

from datatrove.pipeline.readers import ParquetReader

# limit determines how many documents will be streamed (remove for all)
# this will fetch the Portuguese filtered data
data_reader = ParquetReader("hf://datasets/HuggingFaceFW/fineweb-2/data/por_Latn/train", limit=1000) 
for document in data_reader():
    # do something with document
    print(document)

###############################    
# OR for a processing pipeline:
###############################

from datatrove.executor import LocalPipelineExecutor
from datatrove.pipeline.readers import ParquetReader
from datatrove.pipeline.filters import LambdaFilter
from datatrove.pipeline.writers import JsonlWriter

pipeline_exec = LocalPipelineExecutor(
    pipeline=[
        ParquetReader("hf://datasets/HuggingFaceFW/fineweb-2/data/por_Latn/train", limit=1000),
        LambdaFilter(lambda doc: "hugging" in doc.text),
        JsonlWriter("some-output-path")
    ],
    tasks=10
)
pipeline_exec.run()

Using huggingface_hub

from huggingface_hub import snapshot_download
folder = snapshot_download(
                "HuggingFaceFW/fineweb-2", 
                repo_type="dataset",
                local_dir="./fineweb2/",
                # download the Czech filtered + removed data
                allow_patterns=["data/ces_Latn/train/*", "data/ces_Latn_removed/train/*"])

For faster downloads, make sure to install pip install huggingface_hub[hf_transfer] and set the environment variable HF_HUB_ENABLE_HF_TRANSFER=1.

Using datasets

As mentioned above, load_dataset will not work for und_ or _removed splits.

from datasets import load_dataset
# get Croatian data
fw = load_dataset("HuggingFaceFW/fineweb-2", name="hrv_Latn", split="train", streaming=True)

Dataset processing steps

We used the 🏭 datatrove library to process the data. You can find a working script that launches the entire processing pipeline here.

The processing pipeline had to be heavily adapted for a multilingual setting. As each language has its own peculiarities, we individually tuned each filter, defining different thresholds and stopwords for each language. 📊 These thresholds and stopwords are available in /configs/{iso3_lang}_{script}.yml in our github repo.

The starting point for our dataset was the non-English data (< 0.65 score in English) we obtained when processing the original FineWeb. This data was text extracted using trafilatura and went through our URL filters (for more info see 🍷 FineWeb. To this data, we applied the following processing steps:

1. Additional Language Identification and filtering 🔍
2. Deduplication per language 🔄
3. Filtering per language 🧹
4. PII Anonymization and fixes 🎭

Language Identification 🌍

Performed using GlotLID, which not only covers a wider variety of languages (2000+ available labels) compared to fasttext176 (used in the original FineWeb), as it also identifies the script used in each document. 📜

For each language, we defined different minimum language classifier confidence scores to keep a document.

Deduplication 🗃️

Unlike in 🍷 FineWeb, where data was deduplicated per CommonCrawl snapshot, in 🥂 FineWeb2, data is deduplicated per language, globally. However, following our deduplication findings in the original 🍷 FineWeb, while we remove all except one document from each duplicate cluster, we save the size of this cluster in the kept document's metadata, saved in minhash_cluster_size. This allows us to "re-hydrate" the dataset: by upsampling documents based on their cluster size, we see clear performance improvements for some languages, particularly high resource ones. 📈

We think upsampling weights should be dataset specific, and have therefore used the filtering rates of each duplicate cluster to compute different weights per language. They are available on our Github repo, along with sample code to Rehydrate the dataset.

WARNING: If you do not upsample based on these weights, dataset performance may be lower than the one obtained on our results.

Data Filtering 🧹

We mostly kept the original 🍷 FineWeb set of filters, and do not create new filters targeting individual languages. As such, we had to extensively ablate on different processes of adapting the English filters to all the languages we supported. 🔍

Based on the results of our experiments, we also disabled/changed global values of some specific filters:

• For FineWebQuality filters, we removed short_line_thr and changed char_dup_ratio from 0.01 to 0.1.
• Gopher Repetition filter: disabled paragraph related filters as trafilatura does not keep them ❌
• C4 filters: we did not include the C4 filters as they seemed to degrade performance in this multilingual setting 📉

PII Anonymization and fixes 🎭

• PII Removal: Kept unchanged, emails and ip addresses are anonymized. ✉️
• We applied FTFY to fix encoding issues. 🔧
• Added some code to fix trafilatura created artifacts related to tables 🛠️

We will soon release more details regarding the reasoning behind each of these decisions in our upcoming blogpost.

Dataset performance evaluation and ablations

We chose 9 diverse (in script, language family and resource availability) languages for our ablation setup: Chinese, French, Arabic, Russian, Thai, Hindi, Turkish, Swahili, and Telugu. We then selected high signal tasks for these languages out of almost 200 benchmarks. We wrote an entire blogpost about this process: FineTasks, where you will find the full list of tasks we evaluated on, as well as how they were selected. As for metrics, we use normalized probability mass (not accuracies!) for discriminative tasks and f1 for generative tasks, as these metrics have proven to be far more stable than their alternatives.

We conducted our dataset performance ablations and evaluations by training a series of 1.45B parameters models on ~30 billion tokens, tokenized using the gemma tokenizer. To compare 🥂 FineWeb2 with other datasets, we also trained one of these 1.45B models per target dataset, on 30 billion tokens sampled from it (or the entire dataset when its size was < 30 billion tokens). We chose 30B as some of the comparison datasets were relatively small for some languages, but we will soon release some longer ablation runs.

Hyper-parameters for ablation models

The detailed configurations for training the models can be found here.

Comparison with other datasets

Note: the results below use an older version of the dataset. Please check our paper for updated results. You will find all the evaluation results in the repo files. The 🥂 FineWeb2 runs were trained on the final data (dedup+filtering) with re-hydration (see the section on deduplication above), unless explicitly stated (e.g. Swahili).

We compared 🥂 FineWeb2 with the following multilingual datasets:

• mC4
• CC-100
• HPLT v1.2
• CulturaX
• HPLT V2.0

And with language specific monolingual datasets:

• ArabicWeb24 (arabic)
• Arabic-101B (arabic)
• Croissant (french)
• Sangraha (hindi & telugu)
• Odaigen(hindi)
• Omnia Russica (russian)
• Sea CommonCrawl (thai)
• VNGRS-Web-Corpus (turkish)
• MNBVC (chinese)
• TigerBot (chinese)
• MAP-CC (chinese)

Expand each individual language to see the corresponding plot. The error bars correspond to one standard deviation of the scores of 4 models trained on different randomly sampled 30B tokens of unfiltered CommonCrawl data.

Arabic

French

Hindi

Russian

Swahili

For Swahili, the filtered data (around ~1B tokens) performs worse than the deduplicated (filtered+removed subsets) data (around ~3B tokens). We believe this is due to the small number of remaining tokens.

Telugu

Thai

Turkish

Chinese

TigerBot and MAP-CC outperform 🥂 FineWeb2, possibly due to filters specificaly targeting Chinese.

Dataset card for 🥂 FineWeb2

Dataset Summary

This dataset was created by processing 96 CommonCrawl dumps comprising web data crawled from the summer of 2013 to April 2024. 🥂 FineWeb2 includes a variety of domains and topics in a variety of languages and is primarily intended to be used as a research artifact on public data in the context of pretraining datasets for large language models. The CommonCrawl data was carefully processed, deduplicated and filtered with the 🏭 datatrove library, resulting in the largest publicly available multilingual clean LLM pretraining dataset.

Dataset Structure

Data Instances

The following is an example sample from the dataset. It is part of the French (fra_Latn) data, originally belonged to the CC-MAIN-2013-20 CommonCrawl snapshot and was crawled on 2013-05-19T07:12:36Z.

{
   "text": "Il y a 61 ans le match le plus long de l'histoire\nLe 6 janvier 1951 les Rochester Royals recevaient les Indianapolis Olympians pour ce qui allait être le match le plus long de l'histoire. Rochester qui sortait d'une victoire face aux Knicks de New York en prolongation étaient sur une série de 7 victoires avant la réception d'Indianapolis. Au final un match remporté au bout de la nuit par les Olympians en 6 prolongations et un tout petit score de 75 à 73. les équipes n'avaient shooté que 23 fois au total des 6 prolongations! (l'horloge de tir n'était pas encore utilisée)\nCe match reste à ce jour le plus long de l'histoire avec 78 minutes de jeu.",
   "id": "<urn:uuid:5013b1b9-5092-40f8-8d79-c517970dd814>",
   "dump": "CC-MAIN-2013-20",
   "url": "http://basket-infos.com/2012/01/06/il-y-a-61-ans-le-match-le-plus-long-de-lhistoire/",
   "date": "2013-05-19T07:12:36Z",
   "file_path": "s3://commoncrawl/crawl-data/CC-MAIN-2013-20/segments/1368696384213/warc/CC-MAIN-20130516092624-00033-ip-10-60-113-184.ec2.internal.warc.gz",
   "language": "fra",
   "language_script": "Latn",
   "language_score": 0.9994362592697144,
   "minhash_cluster_size": 1,
   "top_langs": "{\"fra_Latn_score\": 0.9994362592697144}"
}

Data Fields

• text (string): the main text content
• id (string): original unique identifier for this sample from CommonCrawl
• dump (string): the CommonCrawl dump this sample was a part of
• url (string): url to the original page where text was present
• date (string): crawl date (from CommonCrawl)
• file_path (string): s3 path for the individual CommonCrawl warc file containing this sample
• language (string): ISO 639-3 code for the language of this sample
• language_script (string): script of the text, for example Latn
• language_score (float): language prediction score as reported by the GlotLID classifier
• top_langs: language-script pairs for which the language classifier
• minhash_cluster_size: number of samples in the minhash cluster of this sample. See the deduplication section to learn why this might be useful

Data Splits

See "Languages and available subsets" above.

Dataset Creation

Curation Rationale

While multiple open-weights models have regularly been released in recent months, these releases often do not include the model's training data. With 🥂 FineWeb2 we aim to provide the open source community with a very large clean pretraining dataset that can be used to push the envelope on truly open source models (open source models where data is also released). We also seek to improve the representation of lower resource (and often ignored) languages, and deliberately chose a language classifier that supported a large number of language labels.

Source Data

The source data consists of webpages crawled by the CommonCrawl foundation over the 2013-2024 time period.

We then extracted the main page text from the html of each webpage, identified its language, deduplicated the data per language and then filtered with specific thresholds adapted to each language.

Data processing steps

See "Dataset processing steps" above.

Annotations

We augment the original samples with the language, language_script, language_score, top_langs and minhash_cluster_size annotations. The language related annotations are automatically generated by our language filter. minhash_cluster_size is computed during the deduplication process, by saving the size of each duplicate cluster before removing all of its documents except one.

Personal and Sensitive Information and opt-out

We anonymize email addresses and public IP addresses.

For emails, we apply a regex pattern and replace any occurrence of an email address with either email@example.com or firstname.lastname@example.org. For IP addresses, we also employ a regex pattern and then further filter to only anonymize IP addresses allocated for public networks. Matched IP addresses are then replaced with one of the following randomly generated IP addresses, which at the time of dataset creation were not responding to ping requests: 22.214.171.124, 126.96.36.199, 188.8.131.52, 184.108.40.206, 220.127.116.11, and 18.104.22.168. We decided against applying regex patterns for phone numbers due to the high false positive rate.

Despite our efforts, given that 🥂 FineWeb2 is sourced from the internet at large, it is very likely that some personable identifiable information (PII) will be present. If you find your own PII in 🥂 FineWeb2 and would like it removed, please fill out our PII removal/opt out form.

CommonCrawl respects robots.txt at crawl time, but if you are a webmaster and find your website in 🥂 FineWeb2 and would like to have it removed, you may also use the PII removal/opt out form.

Considerations for Using the Data

Social Impact of Dataset

With the release of this dataset we aim to make model training more accessible to the machine learning community at large.

While multiple open-weights models with strong performance have been publicly released in the past, more often than not these releases are not accompanied by the corresponding training dataset. This is unfortunate as the dataset specificities and characteristics have been demonstrated to have a very large impact and role in the performances of the models. As the creation of a high quality training dataset is a fundamental requirement to training an LLM capable of excelling at downstream tasks, with 🥂 FineWeb2 we (a) not only make the dataset creation process more transparent, by sharing our entire processing setup including the codebase used, we also (b) help alleviate the costs of dataset curation, both in time and in compute, for model creators by publicly releasing our dataset with the community.

While LLM advancements have primarily focused on English, Chinese, and other Western languages, this release prioritizes broader language support. We consulted with practitioners who develop LLMs for diverse languages to address their specific requirements, such as proper word segmentation (particularly for scripts that don't use whitespace separation) and handling language-specific punctuation, ensuring that medium and lower resource languages were not an afterthought.

Discussion of Biases

Efforts were made to minimize the amount of NSFW and toxic content present in the dataset by employing filtering on the URL level. However, there are still a significant number of documents present in the final dataset that could be considered toxic or contain harmful content. As 🥂 FineWeb2 was sourced from the web as a whole, any harmful biases typically present in it may be reproduced on our dataset.

Some filters might disproportionately target specific domains. One such example is poetry: we noticed that the punctuation filter removes a lot of poems.

We deliberately avoided using machine learning filtering methods that define text quality based on the similarity to a “gold” source such as wikipedia or toxicity classifiers as these methods have been known to disproportionately remove content in specific dialects and overclassify as toxic text related to specific social identities, respectively.

Other Known Limitations

While the language classifier we used, GlotLID supports over 2000 language labels, its performance is not ideal for all of them. The training data for many languages is hard to obtain and, additionally, the classifier is prone to sometimes mistaking closely related languages (for instance, Standard Arabic and Arabic dialects or Croatian and Bosnian). We tried to mitigate this by curating stopwords for each language, but these might also not be effective in all cases.

Due to resource constraints and limited access to native speakers, we couldn't test each language individually. We encourage users to review our filtering approach for their languages of interest and modify the processing if needed. To support this, we've made available all data removed by our filtering pipeline (see "Languages and available subsets" above for more info).

You should also probably consider complementing 🥂 FineWeb2 with specialized curated sources (such as Wikipedia, for example) as they will likely have better formatting than the wikipedia content included in 🥂 FineWeb2 (we did not tailor the processing to individual websites).

Additional Information

Licensing Information

The dataset is released under the Open Data Commons Attribution License (ODC-By) v1.0 license. The use of this dataset is also subject to CommonCrawl's Terms of Use.

Citation Information

@misc{penedo2025fineweb2pipelinescale,
  title={FineWeb2: One Pipeline to Scale Them All -- Adapting Pre-Training Data Processing to Every Language}, 
  author={Guilherme Penedo and Hynek Kydlíček and Vinko Sabolčec and Bettina Messmer and Negar Foroutan and Amir Hossein Kargaran and Colin Raffel and Martin Jaggi and Leandro Von Werra and Thomas Wolf},
  year={2025},
  eprint={2506.20920},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2506.20920}, 
}