OpenAI's CEO, Sam Altman, revealed that the company is preparing to introduce its first open-weight AI model since GPT-2. This new model offers a balance between open-source and proprietary systems. OpenAI raised $40 billion, including $30 billion from SoftBank, on the same day as the announcement.

Altman mentioned that they have been contemplating this move for a while but had other priorities. The company is organizing developer events to gather feedback and test early prototypes before finalizing decisions. The open-weight AI model shares pre-trained parameters of a neural network, allowing developers to use it for inference and fine-tuning while withholding critical development details.

During a Reddit AMA session in February, Altman acknowledged OpenAI's past preference for closed-source development and the ongoing internal discussions about embracing open-source practices. Additionally, Kevin Weil, OpenAI's chief product officer, mentioned the possibility of open-sourcing older models.

On another front, OpenAI has faced concerns about technology misappropriation by DeepSeek, a Chinese AI startup, known for its completely open-source approach with the R1 AI model. OpenAI recently launched GPT-4.5 (Orion) and plans to introduce GPT-5 in May. The company is a part of the Stargate Project, a massive initiative to construct data centers approved by President Donald Trump. With the closure of a substantial funding round, including $30 billion from SoftBank, OpenAI's valuation now stands at $300 billion.