OpenAI releases two open-weight AI models, including one that runs well on Apple Silicon Macs

Living up to its name, OpenAI has released not one but two new open-weight AI models after promising to deliver a new open-weight model earlier this year. The two models, gpt-oss-20b and gpt-oss-120b, are available to download for free now.

What makes open-weight models special? Specifically, these are AI models that can be downloaded and run on computers with adequate resources for powering local AI models. No internet connection is required because access to the model provider’s server isn’t involved. This also allows developers to build custom tools using the open AI models.

OpenAI describes the ’20b’ variant as a medium-sized open model while it calls the ‘120b’ variant a large-sized open model for running on “most desktops and laptops.”

OpenAI says the smaller model works best with at least 16GB VRAM or unified memory and is “perfect for higher-end consumer GPUs and Apple Silicon Macs.” The larger, full-sized model wants at least 60GB VRAM or unified memory.

While these models aren’t the most powerful AI tools from OpenAI, they should satisfy the demand for open-weight models from the team behind ChatGPT.

The rise of DeepSeek out of China last year put pressure on OpenAI to deliver a modern open-weight model. Prior to today, OpenAI hadn’t released an open-weight model since its early days in 2019.

Learn much more about OpenAI’s two new open-weight models here.

FTC: We use income earning auto affiliate links. More.