OpenAI Unveils Offline AI Models for Local Device Use

OpenAI has introduced a significant leap in artificial intelligence accessibility with the release of its latest models, gpt-oss-120b and gpt-oss-20b, capable of running directly on local devices such as smartphones and laptops.

This marks a departure from traditional cloud-based AI, offering a model that functions without a constant internet connection or the need for vast computing infrastructure.

According to CEO Sam Altman, this release represents a major milestone. “This is a big deal,” he stated on X, noting that years of research and billions in investment have led to the development of open-weight models with broad usability.

Unlike conventional models that rely on cloud data centers, these offline tools offer greater data privacy, environmental sustainability, and customization, albeit with the limitation of text-only outputs.

Altman emphasized the empowerment of individuals through AI tools that allow direct control and modification. “Most people will still use ChatGPT, but this gives them an option to run AI themselves when privacy and autonomy matter most,” he said.

To ensure safety, OpenAI partnered with independent security experts to assess risks around misuse, including cybersecurity threats and biohazard scenarios.

This is the first time OpenAI has released an open model of this scale since GPT-2. The gpt-oss models are now available for free on Hugging Face and Databricks, signaling a renewed push to democratize powerful AI.