OpenAI introduces groundbreaking AI models for local use

On August 6, 2025, OpenAI announced the release of two new artificial intelligence models, gpt-oss-120b and gpt-oss-20b, that can operate without an internet connection on consumer devices such as laptops and smartphones. This marks a significant shift from conventional cloud-based AI models, which typically require extensive data centers and resources to function effectively. OpenAI CEO Sam Altman highlighted the environmental benefits of these new models, as they require fewer compute resources, thus lowering their carbon footprint. The motivation behind the development of these locally-run models stems from a desire to enhance individual empowerment by allowing users greater control over their AI tools. Altman stressed the importance of privacy benefits, asserting that users should have the ability to modify and personalize their AI experiences directly. This puts AI in the hands of more people, enabling them to leverage advanced technology for their creative and research purposes. OpenAI aims to drive innovation in the field of artificial intelligence by making these models widely accessible. According to Altman, the new gpt-oss models are expected to spark a significant uptick in innovation and increase productivity in various sectors, allowing more individuals to contribute meaningfully to technological advancements. Although this is the first launch of an open-weight model since GPT-2, concerns around safety and deployment in sensitive areas were addressed through rigorous external testing to mitigate risks. These new models are now available on platforms like Hugging Face and Databricks, allowing developers to download them for free. The introduction of these models symbolizes a critical step toward democratizing access to AI technologies, ensuring that individuals not only benefit from the advantages of artificial intelligence but also have the capability to harness it responsibly.