For the longest time, OpenAI has been known for playing its cards close to the chest when it comes to releasing its most powerful AI models. But now, that might be changing.
In a move that’s stirring curiosity across the AI community, OpenAI is gearing up to release a new open language model this summer its first truly open model since GPT-2. This isn’t just a casual experiment, either. The company seems serious about doing it right, from outperforming benchmarks to making it accessible with minimal restrictions.
Aidan Clark, OpenAI’s VP of research, is at the helm of this project. Though it’s still in its early days, the model is said to be in the same league as their advanced reasoning models is already generating buzz. The goal? To compete head-to-head with other open models like Meta’s Llama or Google’s Gemma, but without the frustrating usage limitations that have drawn criticism from developers and researchers alike.
What’s even more interesting is how OpenAI plans to license the model. Unlike some competitors, they’re exploring a very permissive license. Think: fewer hoops to jump through for anyone looking to experiment, build, or even commercialize products using the model.
This strategic pivot comes at a time when other players, especially DeepSeek from China, are embracing openness and thriving because of it. DeepSeek’s open approach has led to rapid global adoption and significant investor attention. Meta’s Llama series, which leans into open AI development, reportedly hit over 1 billion downloads as of March. Clearly, the market is hungry for openness.
The upcoming model is designed to be simple to use just “text in, text out.” It’s expected to run on high-end consumer hardware and may even come with an optional “reasoning mode,” similar to what other companies like Anthropic have implemented.
CEO Sam Altman has been vocal lately about rethinking OpenAI’s stance on open source. In a recent Reddit Q&A, he acknowledged that perhaps they’ve been too guarded in the past. He hinted that while not everyone at the company agrees, there’s a need to develop a better open-source strategy one that doesn’t compromise safety but also doesn’t wall off progress.
Speaking of safety, OpenAI plans to thoroughly test and “red team” the model before launch. A full model card essentially a detailed breakdown of how the model was trained, tested, and evaluated will accompany the release. It’s a move to reassure both developers and critics that this open model won’t be a free-for-all, but rather a responsibly launched tool.
Altman admitted that OpenAI’s approach to safety and transparency hasn’t been perfect in the past. But this new direction signals a more thoughtful and community-focused approach to innovation one where openness and responsibility aren’t seen as mutually exclusive.
As we wait for summer, all eyes are on OpenAI to see if they’ll follow through and if this new model might just be the beginning of a more open, collaborative chapter in their story.
