OpenAI is releasing a brand new open-weight mannequin dubbed GPT-OSS that may be downloaded free of charge, be personalized, and even run on a laptop computer.
The mannequin is available in two variants: 120-billion-parameter and 20-billion-parameter variations. The larger model can run on a single Nvidia GPU and performs equally to OpenAI’s current o4-mini mannequin, whereas the smaller model performs equally to o3-mini and runs on simply 16GB of reminiscence. Each mannequin variations are being launched at the moment through platforms like Hugging Face, Databricks, Azure, and AWS underneath the Apache 2.0 license, which permits them to be broadly modified for business functions.
That is OpenAI’s first open-weight mannequin in over six years, years earlier than the debut of ChatGPT. Till earlier this 12 months, CEO Sam Altman cited security considerations as the principle motive for not releasing a follow-up. In the meantime, builders have flocked to open fashions as a consequence of their decrease value and customizability. In January, after the rise of DeepSeek, Altman stated that OpenAI had “been on the improper facet of historical past” by not releasing its personal open fashions.
Now, OpenAI is reasserting itself with an open-weight mannequin that it says can carry out reasoning duties, browse the online, write code, and function brokers through the corporate’s current APIs. “I believe lots of people are literally stunned to know that the overwhelming majority of our prospects are already utilizing a variety of open fashions,” Chris Cook dinner, an OpenAI researcher, stated throughout a media briefing. “We needed to plug that hole and permit them to make use of our expertise throughout the board.”
On the protection entrance, OpenAI says that GPT-OSS is its most rigorously examined mannequin to this point, and that it was examined with exterior security companies to make sure it doesn’t pose dangers in areas like cybersecurity and organic weapons. The mannequin’s chain of thought, or seen course of used to reach at a solution, is proven “to observe mannequin misbehavior, deception and misuse,” in accordance with an organization press launch. Its output is text-only and, like all of OpenAI’s fashions, GPT-OSS’s coaching information is undisclosed.
“The staff actually cooked with this one.”
OpenAI hasn’t shared benchmarks evaluating GPT-OSS to different open fashions like Llama, DeepSeek, or Google’s Gemma. Each variants of GPT-OSS carry out equally to OpenAI’s closed reasoning fashions on coding duties and exams like Humanity’s Final Examination. “These are unimaginable fashions,” stated OpenAI cofounder Greg Brockman. “The staff actually cooked with this one.”
OpenAI isn’t committing to a launch schedule for future variations of GPT-OSS, nevertheless it hopes that the mannequin will probably be utilized by smaller builders and corporations that need extra management over how their information is used. “We’ve at all times believed that in case you decrease the barrier to entry, then innovation simply goes up,” stated Brockman. “You let individuals hack, then they’ll do issues which are extremely stunning.”
