OpenAI has long been trolled for not being ‘open; at all — its models are proprietary and hidden, unlike the open models released by companies like DeepSeek and Meta — but it could be looking to change all that.
OpenAI will release an open weights model in the next few months, OpenAI CEO Sam Altman has announced. “We are excited to release a powerful new open-weight language model with reasoning in the coming months, and we want to talk to devs about how to make it maximally useful,” Altman posted on X. “We are excited to make this a very, very good model!” he added.
“We are planning to release our first open-weigh language model since GPT-2. We’ve been thinking about this for a long time but other priorities took precedence. Now it feels important to do. Before release, we will evaluate this model according out our preparedness framework, like we would for any other model. and we will do extra work given that we know this model will be modified post-release. We still have some decisions to make, so we are hosting developer events to gather feedback and later play with early prototypes. we’ll start in SF in a couple of weeks followed by sessions in Europe and APAC. We’re excited to see what developers build and how large companies and governments use it where they prefer to run a model themselves,” he added.
“We will not do anything silly like saying that you cant use our open model if your service has more than 700 million monthly active users. we want everyone to use it!” Altman added, in a apparent dig at Meta which had similar restrictions on its own open-source models.
“Safety is a core focus of our open-weight model’s development, from pre-training to release,” wrote Johannes Heidecke who works on model safety at OpenAI. “While open models bring unique challenges, we’re guided by our Preparedness Framework and will not release models we believe pose catastrophic risks,” he added.
OpenAI has long been criticized for not open sourcing its models. The company had been founded in 2015 with the aim of creating an “open” alternative to Google, which was then the most dominant player in the AI space. OpenAI had open-sourced some models till GPT-2, the precursor to GPT-3 which served as the basis for the first version of ChatGPT. But ever since ChatGPT emerged as a huge success, OpenAI hasn’t open-sourced any of its popular models. Several competitors, like Meta and Mistral, have partially open-source models, while China’s DeepSeek has released some of the most fully-open source models around.
All this had put public pressure on OpenAI to release some of its own models as open source as well. It made business sense too — open-source models like DeepSeek R1 were comparable in quality to its own, and were far cheaper. As such, OpenAI appears to have decided that it’s time to open-source some of its own models. They won’t likely be frontier models, but the development community would hope that these models are good enough to be used and played around with. And with the biggest player in the space releasing open-source models, it could only accelerate AI progress and development going forward.