Petals (Decentralized Inference) by BigScience is **not abandoned**. It remains an active project. As of June 2026, Petals continues to be maintained and developed as part of the BigScience research workshop. It is actively used for running and fine-tuning large language models (such as Llama 3.1, Mixtral, and Falcon) in a decentralized, BitTorrent-style manner. Key points regarding its current status: * **Active Development:** The project is still being updated, with recent documentation and community activity visible on its official website (petals.dev) and GitHub repository. * **Functionality:** It continues to support modern models like Llama 3.1 (up to 405B) and remains a functional tool for collaborative inference and fine-tuning. * **Community:** The project maintains an active community, including a Discord server for development discussions and contributions. You can find the latest information, documentation, and ways to contribute your GPU resources at [petals.dev](https://petals.dev/) or on their [GitHub repository](https://github.com/bigscience-workshop/petals).