OSCAR-2301
OSCAR ProjectOSCAR-2301 is a multilingual web corpus extracted from the November/December 2022 Common Crawl snapshot and distributed by language in original and deduplicated versions, covering 151 languages. Its documents carry KenLM-based adult content detection and precomputed locality-sensitive hashes for near deduplication. The OSCAR project publishes it.
Openness
3 high confidence- license
- cc0-1.0(metadata)
- gated
- manual(login + terms acceptance)
Manually gated and the card states access is temporarily suspended pending clarification.
- https://huggingface.co/api/datasets/oscar-corpus/OSCAR-2301 recorded 2026-09-29
Hub API record lists the dataset as gated manual with license cc0-1.0, and its card notice reads "We are forced to temporarily suspend access to OSCAR".
- https://huggingface.co/datasets/oscar-corpus/OSCAR-2301 recorded 2026-09-29
Dataset card gating notice reads "We are forced to temporarily suspend access to OSCAR" and says access is on "manual control" until the situation is clarified; the metadata keywords list "cc0-1.0".
Adoption
1 high confidenceAdoption is measured by Hugging Face downloads of oscar-corpus/OSCAR-2301, the only usage channel on record for OSCAR-2301.
- https://huggingface.co/api/datasets/oscar-corpus/OSCAR-2301 recorded 2026-10-07
663 downloads in the trailing 30 days for oscar-corpus/OSCAR-2301
Capability
3 high confidenceColossal OSCAR, which extends OSCAR 23.01 to ten monthly Common Crawl snapshots, supplies about 53% of the pretraining tokens of BSC's Salamandra models, and ALIA also trains on it. Access to the raw OSCAR-2301 corpus is suspended, and FineWeb-2 reports outperforming multilingual web corpora such as CulturaX, which itself includes OSCAR 23.01.
- https://huggingface.co/BSC-LT/salamandra-7b recorded 2026-09-29
Model card states "The pretraining corpus is predominantly composed of data from Colossal OSCAR, which contributes a significant 53.05% of the total tokens."
- https://huggingface.co/datasets/HuggingFaceFW/fineweb-2 recorded 2026-09-29
FineWeb2 card says it "outperforms other popular pretraining datasets covering multiple languages" such as CC-100, mC4, CulturaX or HPLT; OSCAR is not named.
- https://huggingface.co/datasets/oscar-corpus/colossal-oscar-1.0 recorded 2026-09-29
Colossal OSCAR card says it "currently contains all the features present in OSCAR 23.01, the main difference being its size".
- https://huggingface.co/datasets/uonlp/CulturaX recorded 2026-09-29
CulturaX card says it combines mC4 with "all accessible OSCAR corpora up to the present year, including 20.19, 21.09, 22.01, and 23.01".
Verified 2026-09-29, except some scores, still awaiting re-confirmation