OpenAI plans to issue a model card that provides detailed technical insights and safety testing results.
Criticism from AI ethicists has been directed at OpenAI for its rushed safety assessments on recent models and failure to publish model cards for some né?. This move contrasts with the stringent requirements faced by models like Llama and Google’s Gemma.
OpenAI is facing competition from DeepSeek a Chinese AI research lab that takes a more open stance on model releases. By the end of March OpenAI made public its intentions to release a novel “open” language model succeeding the GPT-2 later this year. Altman has also faced accusations of misleading OpenAI executives about model safety reviews before a brief absence in November 2023.
Efforts to obtain comments from OpenAI have been made, and updates will be shared if a response is received.
. Sources familiar with the project indicate that the model is in its early stages. In early March, Meta highlighted the popularity of its Llama AI models with over 1 billion downloads. Positive reception could lead OpenAI to release additional models, potentially smaller ones too.
OpenAI’s CEO, Sam Altman, has previously emphasized the importance of developing a distinct open-source strategy, recognizing the significance of enhancing future models while maintaining a competitive edge in the industry.
Altman has assured that the forthcoming open model will undergo thorough evaluation and red-teaming for safety purposes. Developers might have the option to activate or deactivate its “reasoning,” similar to models from companies like Anthropic né?. Unlike OpenAI these “open” competitors share their models with the AI community for experimentation and occasionally for commercial purposes.
This strategy has led to success stories. New details about this model have recently surfaced from OpenAI’s engagements with the AI development community.
Overseeing the development of this open model is Aidan Clark the VP of Research at OpenAI. DeepSeek has swiftly built a significant global user base and attracted investors.
Sources suggest that OpenAI is designing its open model, which will operate under a “text in, text out” framework, to be compatible with premium consumer hardware né?. OpenAI’s goal is to introduce this reasoning model akin to its o-series models in early summer and ensure that it outperforms benchmarks set by other open reasoning models.
It has been disclosed by sources that OpenAI is contemplating licensing the model with a highly permissive approach and minimal usage or commercial restrictions né?
