CIOTech Outlook Team | Wednesday, 06 August 2025, 04:39 IST
OpenAI unveiled two open-weight language models, gpt-oss-120b and gpt-oss-20b, designed for advanced reasoning and optimized for local execution on laptops. These models mark OpenAI’s first open-weight release since GPT-2 in 2019, offering developers access to publicly available trained parameters for analysis and fine-tuning without needing the original training data.
Compared to the open-source models that incorporate the entire source code, training data, and procedures, the open-weight models only concentrate on the availability of the weights.
"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," said OpenAI co-founder Greg Brockman during a press briefing. The larger model, gpt-oss-120b, operates efficiently on a single GPU, while the smaller gpt-oss-20b is compact enough to run on personal computers.
Also Read: Dixon, Tech Mahindra partner for AI automation in plants, R&D
OpenAI claims these models perform comparably to its proprietary o3-mini and o4-mini models, excelling in coding, competition math, and health-related queries. The models were trained on a text-only dataset emphasizing science, math, and coding knowledge, though OpenAI did not provide benchmark comparisons with competitors like DeepSeek’s R1 model.
Amazon revealed that OpenAI’s open-weight models are now available on its Bedrock generative AI marketplace within Amazon Web Services (AWS). This integration marks the first time Bedrock offers an OpenAI model. "OpenAI has been developing great models, and we believe that these models are going to be great open-source options, or open-weight model options for customers," said Atul Deo, Bedrock Director of Product. Deo declined to elaborate on contractual details between AWS and OpenAI. This move comes as Amazon faces challenges, with its AWS unit reporting slower growth compared to competitors, contributing to a recent tumble in its shares.
The AI model market is competitive. Meta Llama models had previously dominated before DeepSeek in China introduced a low-cost reasoning model earlier this year. The OpenAI, a company with a valuation of 300 billion dollars and funded by Microsoft, is trying to raise up to 40 billion dollars in funding led by Softbank Group.