OpenAI is making waves in the AI community with the launch of two new models, gpt-oss-120b and gpt-oss-20b. Available for free download and customisation, these models offer developers and researchers a straightforward way to explore and build on advanced AI technology without breaking the bank.
This move reflects OpenAI’s original nonprofit mission and comes at a time when calls for greater transparency in AI are growing louder. “Going back to when we started in 2015, our aim has always been to develop an Artificial General Intelligence (AGI) that benefits everyone,” explained CEO Sam Altman. By releasing an open-weight model, OpenAI lets users inspect and modify the trained parameters—a practice already embraced by peers like Meta and Chinese startup DeepSeek.
Greg Brockman, OpenAI’s president, expressed his enthusiasm: “This is the first time we’re making an open-weight language model available in a long time, and it’s really something special.” Designed to handle everyday tasks such as internet searches and code execution, the models are optimised to run efficiently even on local systems.
The initiative is set to spark fresh research and inspire practical applications, with partnerships already in place with firms like Orange and Snowflake. OpenAI has also taken care to design these models to avoid misuse, ensuring that ethical considerations remain at the forefront of technological progress.
Earlier this year, Altman acknowledged that past strategies hadn’t fully embraced openness. The decision to stick to the organisation’s nonprofit roots—despite temptations to shift to a fully for-profit approach—raised eyebrows among some investors while earning praise from AI safety advocates and even co-founder Elon Musk. Now, with a structure that includes a profit-generating arm under the oversight of a nonprofit board, OpenAI is charting a balanced and responsible course for the future.
