Photo by Anna Shvets on Pexels

OpenAI has launched GPT-OSS, a new open-weight language model family designed to bring powerful AI capabilities to personal computers. Available for free download and customization, GPT-OSS aims to democratize access to advanced AI technology. Two versions are offered: a larger 120-billion-parameter model and a smaller, more efficient 20-billion-parameter variant.

The 120B model, comparable to OpenAI’s existing o4-mini in performance, can operate on a single high-end Nvidia GPU. The smaller 20B model mirrors the performance of o3-mini and boasts a modest 16GB memory footprint, making it accessible to a wider range of users.

GPT-OSS models are distributed under the permissive Apache 2.0 license, enabling broad use and modification. They are readily available through popular platforms such as Hugging Face, Databricks, Azure, and AWS, streamlining the adoption process for developers and researchers.