OpenThoughts-114k
Open ThoughtsSynthetic reasoning dataset of roughly 114,000 examples spanning math, science, code and puzzles, built by curating problem sets, generating reasoning traces with DeepSeek-R1 and then verifying correctness. It was created by the Open Thoughts team and used to train the OpenThinker 7B and 32B models.
Openness
5 high confidence- license
- apache-2.0
- access
- ungated
- dataset_card
- present
Apache-2.0 licensed, ungated, with dataset card, code repo, and paper.
- https://huggingface.co/api/datasets/open-thoughts/OpenThoughts-114k recorded 2026-09-29
The Hub API record lists the tags "license:apache-2.0" and "arxiv:2506.04178" and "size_categories:100K<n<1M" for the dataset, with the gated flag false.
- https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k recorded 2026-09-29
The dataset page shows the card text "Open synthetic reasoning dataset with 114k high-quality examples covering math, science, code, and puzzles".
- https://huggingface.co/datasets/open-thoughts/OpenThoughts-114k/raw/main/README.md recorded 2026-09-29
The raw dataset card declares "license: apache-2.0" and links the repository "https://github.com/open-thoughts/open-thoughts" and the paper "https://arxiv.org/abs/2506.04178".
Adoption
3 high confidenceAdoption is measured by Hugging Face downloads of the dataset over the trailing month, the only usage channel on record for OpenThoughts-114k.
- https://huggingface.co/api/datasets/open-thoughts/OpenThoughts-114k recorded 2026-10-07
80,882 downloads in the trailing 30 days for open-thoughts/OpenThoughts-114k
Capability
3 high confidenceOpenThinker-32B, trained on this dataset, scored 90.6 on MATH500 and 61.6 on GPQA-Diamond, ahead of the open-data models LIMO-32B, s1-32B and s1.1-32B. The Open Thoughts team has since superseded it with OpenThoughts3-1.2M, which its own blog calls the state-of-the-art open-data reasoning dataset.
- https://huggingface.co/open-thoughts/OpenThinker-32B/raw/main/README.md recorded 2026-09-29
The OpenThinker-32B card compares LIMO-32B, s1-32B and s1.1-32B with OpenThinker-32B, listing "|LIMO-32B|0.8k|56.7|49.3|86.6|58.1|60.0|" against "**90.6**|**61.6**".
- https://www.openthoughts.ai/blog/ot3 recorded 2026-09-29
The OpenThoughts3 post states "are the SOTA open-data model and SFT dataset for reasoning models, respectively" and describes scaling "from Bespoke-Stratos-17k to OpenThoughts-114k and then again to OpenThoughts2-1M".
Verified 2026-09-29, except some scores, still awaiting re-confirmation