AsianFin -- OpenAI has released two new open-source language models—GPT-oss-120b and GPT-oss-20b—marking its first major open-source model release since GPT-2. The models are designed for efficient on-device inference, enabling them to run locally on personal computers and even scaled-down versions on smartphones.

OpenAI CEO Sam Altman described the release as a “major breakthrough,” stating on social media platform X, “GPT-oss is OpenAI’s most advanced open-source model to date, delivering real-world performance comparable to o4-mini, while being fully capable of running on your own computer—or a smaller version on your phone.”

Altman also hinted that OpenAI will unveil “many new things” in the coming days, fueling anticipation around the company’s next wave of AI product announcements.