Media Contact
OpenAI is planning to introduce its latest open-weight language model in the coming months, marking the first release of its kind since the GPT-2 model.
David Widder, a postdoctoral fellow at Cornell Tech, examines ethical, political and economic aspects of ‘open’ artificial intelligence.
Widder says:
“OpenAI has lost the performance edge it once had against competitors’ models. This ‘flattening’ in terms of relative performance means that developers no longer have a clear reason to choose closed OpenAI models, especially when competitive open weight alternatives like DeepSeek are available.
“Given that OpenAI is finding it challenging to make headlines for flashy benchmark-busting models, this change towards releasing an open-weight model appears to be an attempt to compete on criteria beyond performance, and the accompanying world tour to ‘gather feedback’ seems like an attempt to regain the lost hype.
“Aside from this, it is funny when a company called ‘OpenAI’ gets headlines for merely considering releasing an open weight model in the future. As we’ve noted in previous research, even open-weight models do not negate the fact that immense resources are needed to build AI systems or use them at scale, and that these resources are concentrated in the hands of big tech companies.”