The Stack v3: a code dataset with 5 trillion tokens
The Stack v3 contains about 5 trillion tokens of code across 713 filtered languages and 224 million repositories, based on GitHub collection completed by August 2025.
Source: The Stack v3 lança dataset de código com 5 trilhões de tokens (huggingface.co). Text prepared with AI from this source.
What happened and what to do
The Stack v3 presents a dataset with about 5 trillion tokens of code across 713 filtered languages and 224 million repositories. It includes source code directly and is based on a new GitHub collection completed by August 2025. The material identifies model training and fine-tuning as possible uses.
A company considering training or fine-tuning models with code data could implement a pipeline for ingestion, filtering, provenance controls, and data preparation, alongside evaluations to measure model behavior on internal tasks. Before use, it should also define governance criteria and review rights applicable to the dataset and generated outputs.
How the consultancy can help
Wendelmaques can diagnose the opportunity and risks, scope a data pipeline and evaluations, and implement and operate the solution on the company’s own infrastructure.
Next step
Send a short description of your use case to receive a scoped proposal for diagnosis and implementation.
Let's assess your project
I reply with scope and a proposal.