OpenAI has made a significant move in artificial intelligence by releasing two new open-weight language models, namely gpt-oss-120b and gpt-oss-20b, marking its first release since the introduction of GPT-2 in 2019. These text-only models are built to provide more economical alternatives that developers, researchers, and businesses can easily operate and customize.
Open-weight models have their parameters publicly accessible, which enhances transparency and control in AI development. However, it is important to differentiate these from open-source models, which make the complete source code available for modification and use.
This release is part of a growing trend, as multiple technology companies, such as Meta, Mistral AI (backed by Microsoft), and the Chinese startup DeepSeek, have also launched their open-weight models in recent years.
OpenAI President Greg Brockman expressed enthusiasm about the evolving ecosystem in the AI space, emphasizing the company’s commitment to innovation and collaboration. The development involved partnerships with major tech firms like Nvidia, Advanced Micro Devices, Cerebras, and Groq, ensuring the models are compatible with a range of chips.
Nvidia CEO Jensen Huang highlighted OpenAI’s role in showcasing the capabilities of Nvidia AI and commended the advancements in open-source software that are being propelled by these new models.
This release signifies an important step towards greater accessibility in AI technologies, potentially fostering further innovation and research in the field. It presents a hopeful opportunity for developers and researchers to push the boundaries of what’s possible in artificial intelligence.