February 18th.OpenAI CEO Sam Altman posted on X social media platforms on February 17, local time, regarding the company's nextOpen SourceThe direction of the project was solicited for public comment. He suggested two possible directions for open source modeling:One is the "o3-mini class model that is quite small but still needs to run on the GPU" and the other is the "phone-sized model that is optimized as much as possible".This may mean that OpenAI is about to open source a new big model. This may mean that OpenAI is about to open source a new big model, and will be the first time OpenAI has open sourced a big model since OpenAI open sourced GPT-2 in 2019.
1AI notes that when asked in a Reddit "Ask Me Anything" event on January 31 whether OpenAI would make some of its AI modeling technology public and release more research on how the system works, Altman said that the company was discussing the possibility internally and reflected on OpenAI's past stance on open source. He acknowledged that OpenAI had previously taken a position on open source. He acknowledged that OpenAI may have been "on the wrong side of history" in its open source strategy and said that a different open source strategy was needed. He also noted, however, that not everyone in the company shares this view.
Upon its founding in 2015, OpenAI had stated that it would release research and data from its models when it was in the public interest to do so. But the company has since shifted to the development of proprietary models due to competitive pressures and concerns about the security risks that disclosure could pose.
However, recent changes in the AI tech industry seem to have prompted OpenAI to revisit its open source strategy, and DeepSeek's low-cost and open source model has quickly attracted many users, with its large model DeepSeek-R1 trained in just two months and costing only $5.5 million, compared to the cost of similar projects from companies like OpenAI. compared to similar projects from OpenAI and others. Since the launch of ChatGPT at the end of 2022, many AI companies have chosen not to disclose too many technical details in order to remain competitive. But now, with the rise of open-source technologies like DeepSeek, some developers and customers may be turning to more cost-effective open-source solutions.
