Google’s September 30 announcement of Gemini 4 Argon, its new frontier large language model optimized for long-horizon reasoning, coding, and enterprise tasks, drives current trader focus on its Humanity’s Last Exam (HLE) debut. Argon already posts a 57.1% score on the expert-level academic benchmark and leads or ties 12 of 18 internal evaluations, including DeepSWE v1.1 at 77.9% and the Vals Index. It trails top Claude and GPT-6 variants on some public HLE and HLE-Diamond leaderboards while offering a 1-million-token output window and gated rollout via Google’s Fairwind cyber-defense program. Wider API access and any official HLE update could quickly shift implied probabilities as competitive gaps narrow or widen.