OpenAI’s latest frontier models, including the GPT-5 series and GPT-5.6 Sol released in mid-2026, deliver strong gains in agentic coding, multi-step reasoning, and tool use yet still fall short of AGI benchmarks across cognitive domains. Official roadmaps project only incremental scientific discoveries this year, with more substantial advances expected from 2028 onward, while independent evaluations place current systems well below human-level generality. The four-plus months remaining in 2026 offer limited runway for a verified breakthrough or formal announcement that would satisfy the independent panel criteria tied to OpenAI’s Microsoft agreement. Traders therefore assign just 9 percent implied probability to an AGI declaration before 2027, though an unforeseen scaling leap or internal redefinition of the term could still alter the outcome.