The Stack v2

license

permissive(SPDX)

gated

terms-acceptance+contact-info

dataset_card

yes

Requires acceptance of terms and sharing contact info before download, so access is gated.

• https://huggingface.co/api/datasets/bigcode/the-stack-v2 recorded 2026-10-08

  Hub metadata: "gated":"auto" with the gate field "I have read the License and agree with its terms".

• https://huggingface.co/datasets/bigcode/the-stack-v2 recorded 2026-10-08

  Dataset page gated on terms: "You need to agree to share your contact information to access this dataset". The card says files are "licensed with one of the permissive licenses" or carry no license, and documents the dataset in full.

Adoption is measured by Hugging Face downloads of the dataset over the trailing month, the only usage channel on record for The Stack v2.

• https://huggingface.co/api/datasets/bigcode/the-stack-v2 recorded 2026-10-07

  5,571 downloads in the trailing 30 days for bigcode/the-stack-v2

The Stack v2 trained StarCoder2, is the code base behind SmolLM3, and is the source dataset for Stack-Edu. It spans 619 programming languages and is four times the size of the original StarCoder dataset, but as raw code it loses to its own education-filtered slice, Stack-Edu, in ablations.

• https://arxiv.org/abs/2402.19173 recorded 2026-09-29

  The StarCoder2 paper abstract says the dataset "results in a training set that is 4x larger than the first StarCoder dataset" and that the repositories span "619 programming languages".

• https://huggingface.co/blog/smollm3 recorded 2026-09-29

  The SmolLM3 blog lists a stage 1 mix with "Code: 12% - The Stack v2 (16 programming languages)" among its sources.