OpenAI is preparing to release what would be its first fully open-source large language model, according to comments from AI entrepreneur Emad Mostaque and other industry figures discussing the project this week. Mostaque said the system is expected to deliver performance comparable to OpenAI’s proprietary O3-mini model while running locally on a smartphone, a capability that could handle roughly 90 percent of everyday knowledge tasks at little or no cost to users. The move would mark a significant shift for OpenAI, which has so far limited public access to its most powerful models. An open-weight release optimized for mobile devices could broaden the company’s reach among developers and hardware makers seeking to embed generative-AI features without relying on cloud processing. OpenAI has not formally confirmed the timing or specifications of the model.