A public commit to an open-source biology-benchmarks repository dated 13 July introduced an entry labelled “gpt-5-reasoning-alpha,” the first verifiable reference to OpenAI’s next major language model. The file replaced an earlier listing for the company’s “o3” system, sparking industry scrutiny that the successor to GPT-4 is undergoing external evaluation. Researchers and independent developers who located the commit say internal safety reviews appear to be the final hurdle, with several sources forecasting a release in the next two weeks. Documentation circulating in developer circles describes GPT-5 as a router-style system that shifts user queries among specialised reasoning, general and tool-using sub-models, a configuration intended to improve performance while reducing compute costs. OpenAI researcher Jerry Tworek told community forums that GPT-5 will launch without the recent “IMO-winning” reasoning breakthrough, which is slated for a later upgrade toward year-end. The comment reinforces expectations of an imminent rollout while hinting at a rapid development pipeline that already includes work on GPT-6.