OpenAI is preparing to launch its first open AI language model since GPT-2, with plans to debut it in the coming months. The company recently shared a feedback form, asking developers, researchers and the public for input on what they’d like to see in an open-weight model.
OpenAI’s Shift Toward Openness
For years, OpenAI has kept most of its models proprietary, but rivals like Meta (with Llama) and China’s DeepSeek have gained traction by offering open models. Meta’s Llama, for example, has surpassed 1 billion downloads, proving the demand for accessible AI tools.
Now, OpenAI seems to be reconsidering its strategy. CEO Sam Altman admitted in a Reddit Q&A that the company may have been “on the wrong side of history” by not open sourcing more models. He hinted at a new open-source approach, though he clarified it’s not the company’s top priority yet.
What We Know About the New Model
Altman shared on X that the upcoming model will have reasoning abilities similar to OpenAI’s o3-mini. Before release, OpenAI will rigorously test it under its safety framework. The company also plans to host developer events in San Francisco, Europe, and Asia-Pacific to gather feedback and showcase prototypes.
Why This Matters
An open model from OpenAI could accelerate AI innovation, allowing more developers to experiment and build new tools. However, it also raises concerns about misuse, as open models can be modified after release. OpenAI says it will take extra precautions to ensure responsible use.
Meanwhile, new reports suggest internal tensions at OpenAI, including allegations that Altman misled executives about AI safety before his brief removal in late 2023.
This move could mark a major shift in OpenAI’s strategy-balancing openness with safety as competition in AI heats up.
By
