Teknium

license

not-clearly-stated-on-card(no license field in the card YAML, no license section in the card body, no LICENSE file in the repository and no license tag on the Hub API)

card

present

ungated

yes

format

JSON/Parquet

rows

~1M

OpenHermes-2.5 is publicly downloadable and ungated with a full dataset card, but carries no license anywhere the publisher controls: not in the card, not in the repository, and not on the Hub API. Files that download freely without a grant of rights are neither open nor fully closed, which is why this dataset reads as restricted rather than open.

• https://huggingface.co/api/datasets/teknium/OpenHermes-2.5 recorded 2026-09-29

  Hub API record with cardData listing only language, pretty_name and tags, no license tag, and tags "size_categories:1M<n<10M" and "format:json"; siblings are .gitattributes, README.md and openhermes2_5.json, so no LICENSE file. Searched the body for license and LICENSE: none.

• https://huggingface.co/datasets/teknium/OpenHermes-2.5 recorded 2026-09-29

  HF dataset page for teknium/OpenHermes-2.5 rendering the card text "primarily synthetically generated instruction and chat samples" and the title "Dataset Card for Dataset Name", public with no gate notice.

• https://huggingface.co/datasets/teknium/OpenHermes-2.5/raw/main/README.md recorded 2026-09-29

  Raw card whose front matter lists only language, pretty_name and tags, with no license key, and whose body says "reaching 1M, primarily synthetically generated instruction and chat samples". Searched the body for license, licence and terms: none.

Adoption is measured by Hugging Face downloads of teknium/OpenHermes-2.5, the only usage channel on record for OpenHermes 2.5.

• https://huggingface.co/api/datasets/teknium/OpenHermes-2.5 recorded 2026-08-13

  17,420 downloads in the trailing 30 days for teknium/OpenHermes-2.5

The SmolLM2 authors included a 100k-example slice of OpenHermes-2.5 in SmolTalk for its strong knowledge performance on MMLU-Pro. In the same paper's comparison, fine-tuning on OpenHermes-2.5 alone trails the SmolTalk mixture on instruction following and MT-Bench.

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

  SmolLM2 paper section 5.1.4 says "We also added 100k randomly selected OpenHermes2.5 samples due to its strong performance in knowledge (MMLU-Pro)" and Table 10 lists OpenHermes 2.5 at 100k among the SmolTalk components.