Google's Gemini 3.5 Pro is reportedly months behind schedule, with recent training data updates yielding "disappointing" results, particularly in AI coding capabilities. This setback mirrors previous challenges faced by Meta with Llama 4 and xAI with Grok 4, highlighting the difficulty of consistently advancing frontier models.