The gap between open weights LLMs and closed source LLMs

# Summary Open-source LLMs are narrowing the gap with closed-source models, particularly in coding benchmarks, but the trajectory varies significantly by performance metric. While a single benchmark (Artificial Analysis Intelligence Index) suggests open-source models could match closed-source ones by December 2026, analyzing 18 different benchmarks reveals the gap remains relatively stable at around 5 months across most measures, highlighting the difficulty of measuring LLM quality with any single metric.