On September 23, Anthropic rolled out Claude Opus 5.5, a model whose performance across most tasks now rivals that of its premium Fable 5.1. Yet, it operates at a faster pace and lower cost, presenting a formidable challenge to Fable. A mere two hours later, OpenAI followed suit by introducing two new models, GPT-6 Sol and Luna, and implemented significant price reductions to cater to users with varying budgetary constraints. In this latest wave of large model upgrades, price cuts have emerged as a prevailing trend. Nevertheless, under higher reasoning tiers, the quantity of reasoning tokens generated by these models has surged, potentially escalating the actual costs of usage. Concurrently, the performance of certain models has seen a decline rather than an enhancement. Notably, Opus 5.5 has also fine-tuned its approach to handling lengthy and intricate questions. Ultimately, the crux of competition in the current large model market continues to revolve around practical cost-effectiveness.