OpenAI launches two open-weight language models optimised for local use and advanced reasoning
OpenAI releases its first open-weight models since 2019, optimised for laptops and advanced reasoning tasks.
OpenAI has released two new open-weight language models designed to run locally on laptops and deliver performance comparable to its smaller proprietary reasoning models. The launch marks the company’s first open-weight release since GPT-2 in 2019.
Unlike open-source models, open-weight models provide public access to trained parameters but not the complete source code or training data. This allows developers to fine-tune the models for specific tasks without requiring original datasets.
“One of the things that is unique about open models is that people can run them locally. People can run them behind their own firewall, on their own infrastructure,” OpenAI co-founder Greg Brockman said during a press briefing.
The two models, gpt-oss-120b and gpt-oss-20b, are tailored for different hardware capabilities. The larger gpt-oss-120b can run on a single GPU, while the smaller gpt-oss-20b is designed to operate directly on personal computers. Both are said to match the performance of OpenAI’s o3-mini and o4-mini proprietary reasoning models, excelling in coding, competition-level mathematics, and health-related queries.
The models were trained on a text-only dataset with a strong emphasis on science, mathematics, and programming, alongside general knowledge. OpenAI has not published direct benchmark comparisons with other models, such as China’s DeepSeek-R1, which has recently gained attention for its reasoning capabilities.
In a separate announcement, Amazon said OpenAI’s open-weight models are now available through its AWS Bedrock generative AI marketplace. “OpenAI has been developing great models and we believe that these models are going to be great open-weight model options for customers,” said Atul Deo, Bedrock’s director of product, without disclosing the terms of the arrangement.
The release comes as OpenAI, backed by Microsoft and valued at around $300 billion, seeks to raise up to $40 billion in a funding round led by SoftBank Group.