SAN FRANCISCO, U.S. — OpenAI has released two new open-weight language models designed for advanced reasoning tasks and optimized to run on personal computers and laptops, marking a notable shift in its strategy toward greater model accessibility.
The models—gpt-oss-120b and gpt-oss-20b—are the first open models from OpenAI since the release of GPT-2 in 2019.
Unlike open-source models, open-weight models expose the trained weights (parameters) but not the full source code or training datasets. This allows developers to fine-tune or deploy the models locally without needing the original training data.
“One of the things that is unique about open models is that people can run them locally—behind their own firewall, on their own infrastructure,” said OpenAI co-founder Greg Brockman during a press briefing.
Designed for Local Use, Strong in Reasoning and Code
OpenAI says the larger gpt-oss-120b can run on a single GPU, while the smaller gpt-oss-20b is lightweight enough to operate on consumer-grade laptops.
Both models were trained on a text-only dataset focused on general knowledge, science, mathematics, and programming, and are said to match the performance of OpenAI’s proprietary o3-mini and o4-mini models. They excel particularly in coding tasks, competition-level math, and healthcare-related queries.
While the models offer impressive performance, OpenAI did not publish direct benchmark comparisons with rival models such as DeepSeek-R1 or Meta’s Llama models.
OpenAI Models Debut on Amazon Bedrock
In a significant move for cloud-based deployment, Amazon Web Services (AWS) announced that OpenAI’s open-weight models are now available on its Bedrock generative AI marketplace—the first time OpenAI has partnered with Bedrock.
“OpenAI has been developing great models, and we believe these models are going to be great open-weight options for customers,” said Atul Deo, AWS Bedrock’s Director of Product.
Deo declined to reveal any contractual details between Amazon (AMZN.O) and OpenAI.
The announcement comes amid slowing AWS growth and rising competition in the AI cloud space. Meta’s once-dominant Llama models have faced stiff competition in recent months from cost-effective offerings by China’s DeepSeek, while Meta continues to delay Llama 4.
A Competitive Landscape and a New Direction
The release signals a strategic evolution for Microsoft-backed OpenAI, which has traditionally prioritized closed-source offerings.
With the global AI market rapidly expanding and developer demand for customizable, locally-hosted models growing, OpenAI is now re-engaging with the open-weight community.
The company, currently valued at $300 billion, is also reportedly raising up to $40 billion in a funding round led by SoftBank Group (9984.T), signaling continued investor interest despite intensifying competition.
