OpenAI’s decision not to declare AGI despite releasing GPT-6 Astra and related models in September 2026, along with meeting its self-defined “research intern” milestone, underpins the 91% market-implied probability against an announcement before 2027. Current frontier systems demonstrate strong gains in coding, computer use, mathematics, and scientific tasks yet remain specialized rather than broadly autonomous at human-expert level across domains, consistent with Altman’s August 2026 statement that the company was “not quite yet” at AGI. OpenAI’s public roadmap targets a fully automated AI researcher only by March 2028, aligning with slower expert consensus on timelines. A last-minute internal breakthrough or redefinition of AGI criteria could still shift odds, though no verified evidence of such a development has surfaced.