FineMath
Hugging Face'/%3e%3cdefs%3e%3cpattern%20id='pattern0_1240_293949'%20patternUnits='userSpaceOnUse'%20patternTransform='matrix(11.2%200%200%2011.2%203.93617%200)'%20preserveAspectRatio='none'%20viewBox='0%200%2016%2016'%20width='1'%20height='1'%3e%3cg%20id='pattern0_1240_293949_inner'%3e%3crect%20width='16'%20height='16'%20fill='%23f88376'%20mask='url(%23hatch_gaps)'/%3e%3cmask%20id='hatch_gaps'%20maskUnits='userSpaceOnUse'%20x='0'%20y='0'%20width='16'%20height='16'%3e%3crect%20width='16'%20height='16'%20fill='white'/%3e%3crect%20x='6'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='6'%20y='8'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='4'%20y='2'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='4'%20y='10'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='2'%20y='4'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='2'%20y='12'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20y='6'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20y='14'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='14'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='14'%20y='8'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='12'%20y='2'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='12'%20y='10'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='10'%20y='4'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='10'%20y='12'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='8'%20y='6'%20width='2'%20height='2'%20fill='black'/%3e%3crect%20x='8'%20y='14'%20width='2'%20height='2'%20fill='black'/%3e%3c/mask%3e%3c/g%3e%3c/pattern%3e%3c/defs%3e%3c/svg%3e)
FineMath is a classifier-filtered mathematical web corpus extracted from Common Crawl by Hugging Face. It is released in two quality tiers, FineMath-3+ at 34 billion tokens and FineMath-4+ at 9.6 billion tokens, alongside a similarly filtered English portion of InfiMM-WebMath. It is part of the SmolLM2 and SmolLM3 pretraining mixtures and one of the sources in the Dolma 3 pool behind Olmo 3.
Openness
5 high confidence- license
- ODC-BY
- gated
- false
- dataset_card
- present
ODC-BY, ungated, full dataset card.
- https://huggingface.co/api/datasets/HuggingFaceTB/finemath recorded 2026-09-29
The Hub API record lists the dataset files as "finemath-3plus/train-00000-of-00128.parquet", carries the tag "license:odc-by" and reports private and gated as false.
- https://huggingface.co/datasets/HuggingFaceTB/finemath recorded 2026-09-29
The rendered dataset page shows the FineMath card with the sections "Dataset curation" and "Decontamination" plus the line "Use one of the available configs". Searched the body for Request access: none.
- https://huggingface.co/datasets/HuggingFaceTB/finemath/raw/main/README.md recorded 2026-09-29
The raw card front matter reads "license: odc-by" and the body says "The dataset is released in two versions".
Adoption
3 high confidenceAdoption is measured by Hugging Face downloads of HuggingFaceTB/finemath, the only usage channel on record for FineMath.
- https://huggingface.co/api/datasets/HuggingFaceTB/finemath recorded 2026-08-13
29,279 downloads in the trailing 30 days for HuggingFaceTB/finemath
Capability
5 high confidenceFineMath-4+ achieves about a 2x improvement on GSM8K and a 6x improvement on MATH over InfiMM-WebMath in the SmolLM2 annealing ablations, and every FineMath subset outperforms OpenWebMath and InfiMM-WebMath on GSM8K, MATH and MMLU-STEM.
- https://arxiv.org/html/2502.02737v1 recorded 2026-09-29
The SmolLM2 paper reports "FineMath4+ achieves a 2x improvement on GSM8K and a 6x improvement on MATH compared to InfiMM-WebMath" and "All FineMath subsets consistently outperform OWM and InfiMM-WebMath on GSM8K, MATH, and MMLU-STEM".
- https://huggingface.co/datasets/HuggingFaceTB/finemath/raw/main/README.md recorded 2026-09-29
The card reports "FineMath-4+ demonstrates superior performance compared to both FineMath-3+ and InfiWebMath-4+ on GSM8k and MATH" and describes "34B tokens" in FineMath-3+.
Verified 2026-09-27