OpenAI announced two new open-weight AI language models, gpt-oss-120b and gpt-oss-20b, on Tuesday, August 5. These models offer similar capabilities to the company’s O-series, as reported by TechCrunch. This marks the first time since OpenAI rolled out GPT-2 in 2019.
The company stated that the text-only models have been specially designed to be used as lower-cost options, which can be easily run and customized by developers and researchers.
OpenAI releases lower-cost models
OpenAI’s latest release comes after tech giants like Meta and Mistral AI, backed by Microsoft, as well as Chinese startup DeepSeek, unveiled their open-weight models in the past few years.
According to CNBC, the AI model is considered open weight when all its parameters or elements that have been utilized to improve the output and prediction at the time of training are publicly made available. Such models not only offer control and transparency in the respective field, but also remain quite different from the open-source models, where the full source code is made available to customers to modify and use.
OpenAI President Greg Brockman said it is “exciting” to see the developing ecosystem and that the company is looking forward to contributing to it. “…we are excited to contribute to that and really push the frontier and then see what happens from there,” the outlet quoted Brockman as saying.
Also read: Google’s AI tool ‘Big Sleep’ flags 20 security flaws in open-source software without human input
All about OpenAI’s new models
Among the two models, gpt-oss-120b is large and holds the capability to run on a single Nvidia GPU, while the gpt-oss-20b model is the lighter version to run on a consumer laptop having 16GB of memory. The two models, which have been dubbed “state-of-the-art,” remain freely available to download from the Hugging Face platform.
For this project, OpenAI partnered with Nvidia, Advanced Micro Devices, Cerebras, and Groq to see how the models work on various chips. In a statement, Nvidia CEO Jensen Huang said that OpenAI has shown “what could be built on Nvidia AI — and now they’re advancing innovation in open-source software.”
During a briefing, the company highlighted that its new open models are capable of sending complex queries to AI models in the cloud. This means whenever it is unable to perform a specific task, then the developers will be able to connect it to the more capable closed models of the company, per TechCrunch.
The two new AI models, gpt-oss-120b and gpt-oss-20b, underwent a coding test on Codeforces and scored 2622 and 2516. They have outperformed DeepSeek’s R1 model.
The release of OpenAI’s open weight models comes after the company delayed the launch on a repeated basis.
Last month, OpenAI CEO Sam Altman informed in an X post that it required extra time to “run additional safety tests and review high-risk areas.”
FAQs:
1. How to try new OpenAI models?
As per the company, users can download them from Hugging Face and GitHub under the Apache 2.0 license. These are being made available to run on PCs via LM Studio and Ollama.
2. What are the new AI models from OpenAI?
These include open-weight AI language models, gpt-oss-120b and gpt-oss-20b.
3. What’s the price for new OpenAI models?
They are free to download and users can customize them accordingly.