2025-08-05T18:01:37Z
OpenAI CEO Sam Altman announced GPT-oss, his company’s return to releasing an AI language model with open weights.
The company launched two new “open” models, gpt-oss-120b and gpt-oss-20b, on Tuesday.
Altman had previously said OpenAI would release a “very powerful open-source model.”
OpenAI’s AI models are getting more open. At least, some of them are.
OpenAI CEO Sam Altman announced GPT-oss on Tuesday, an “open” family of language models with “open weights” that the CEO said can operate locally on a “high-end laptop” and smartphones.
An AI model with “open weights” is one whose fully trained parameter weights are made publicly downloadable, so anyone can run, inspect, or fine-tune the AI model locally.
“We believe this is the best and most usable open model in the world,” OpenAI CEO Sam Altman wrote on X.
There are two different models: gpt-oss-120b and gpt-oss-20b. The smaller model is designed to run on “most desktops and laptops, ” while the larger model is geared toward higher-end equipment.
Altman said GPT-oss has “strong real-world performance comparable to o4-mini.”
Just before OpenAI’s announcement, rival Anthropic revealed the Claude Opus 4.1.
Tuesday’s announcement was not the long-rumored ChatGPT-5, which could arrive as soon as this week. Instead, the new model is OpenAI’s first open-weight language model since the release of GPT-2 in 2019.
“As part of this, we are quite hopeful that this release will enable new kinds of research and the creation of new kinds of products,” Altman wrote. “We expect a meaningful uptick in the rate of innovation in our field, and for many more people to do important work than were able to before.”
Altman had previously signaled that OpenAI would return to releasing at least some open model, saying that, “We’re going to do a very powerful open source model” that was “better than any current open source model out there.”
Meta has long published the open weights of its LLAMA AI models.