OpenAI CEO Sam Altman recently announced that the company will launch a new open-weighted language model with inference capabilities in the coming months, marking another major advance in the field of open models since GPT-2.

Altman revealed that while the plan was already in the works, its release was repeatedly delayed due to various priorities. Today, OpenAI believes that it is crucial to launch this model that will bring an unprecedented experience to users and developers.

Before the official release, OpenAI will conduct a comprehensive evaluation of the model in accordance with the established framework to ensure its performance and quality. At the same time, knowing that the model may still face adjustments and optimizations after release, OpenAI will carry out additional work to ensure its stability and reliability.

In order to gather more feedback from developers and try out early prototypes, OpenAI plans to hold a series of developer events. The event will start in San Francisco and expand to Europe and Asia-Pacific. Through these activities, OpenAI looks forward to discussing the potential and application prospects of the model with developers.

It is worth noting that, unlike previous OpenAI models, the pre-trained parameters of this open-weighted language model will be shared publicly with the public. This means that developers and researchers can download these parameters, run the model locally, and fine-tune it to their specific needs or incorporate it into a custom application. However, it's important to note that open weights are not the same as fully open source, and the training code, datasets, and verbose logs may remain proprietary or unpublished.

In recent years, the open weight model has become the focus of attention in the industry. Companies such as Meta and Mistral have launched powerful models such as Llama and Mixtral, while new players such as DeepSeek, Alibaba, and Baichuan are also actively launching comparable open-source models. This trend has not only attracted the attention of academic researchers and independent developers, but has also prompted OpenAI to re-examine its model strategy.

With the rise of open weighting models, more and more users are looking for more flexible and customizable model solutions. OpenAI's new open-weighted language model will undoubtedly provide users and developers with more choices and possibilities. We look forward to seeing how this model can inspire more innovation and applications in the future."