OpenWebMath project

license

ODC-BY

gated

false

dataset_card

present

ODC-BY 1.0, ungated, full dataset card.

• https://huggingface.co/datasets/open-web-math/open-web-math recorded 2026-09-29

  The dataset page, public with no access-request gate (its embedded page data reports gated false), renders the card including "OpenWebMath is made available under an ODC-By 1.0 license".

• https://huggingface.co/datasets/open-web-math/open-web-math/raw/main/README.md recorded 2026-09-29

  The card states "OpenWebMath is made available under an ODC-By 1.0 license" and its metadata lists "license: odc-by"; the raw card has no gating block and describes the contents, pipeline and "14.7B tokens" in full.

Adoption is measured by Hugging Face downloads of the dataset over the trailing month, the only usage channel on record for OpenWebMath.

• https://huggingface.co/api/datasets/open-web-math/open-web-math recorded 2026-08-13

  33,285 downloads in the trailing 30 days for open-web-math/open-web-math

OpenWebMath is used in Proof-Pile-2 behind Llemma and in the OLMoE and StarCoder2 math mixes, but newer math corpora beat it in ablations. Every FineMath subset outperforms it on GSM8K, MATH and MMLU-STEM in the SmolLM2 ablations, and Nemotron-CC-Math outperforms it as well.

• https://arxiv.org/html/2502.02737v1 recorded 2026-09-29

  The SmolLM2 paper reports "All FineMath subsets consistently outperform OWM and InfiMM-WebMath on GSM8K, MATH, and MMLU-STEM".

• https://arxiv.org/html/2508.15096v1 recorded 2026-09-29

  The Nemotron-CC-Math paper states "Nemotron-CC-Math-4+ not only surpasses all prior open math datasets" and that variants "consistently outperform or obtain comparable results to OpenWebMath, MegaMath, and FineMath baselines".