Although Anthropic's Mythos has established a new benchmark as the most potent AI large-scale model, it remains inaccessible for public use. At present, the most powerful AI programming large-scale model is still Claude Opus 4.6, which also commands the highest price. Nevertheless, the recent intelligence deterioration of Claude Opus 4.6 has triggered discontent among numerous developers who pay a premium for its services. According to reports, the model's outputs have become less insightful, and it hastily provides modification results, frequently failing at straightforward tasks multiple times. There has been a notable surge in violation prompts that were seldom encountered previously, and token consumption has escalated dramatically. An executive from AMD's AI division also remarked that the intelligence degradation issue has substantially impeded development efforts. Data reveals that the thinking depth of Claude Opus 4.6 has diminished by 67%, accompanied by a sharp decline in code reading frequency, and API costs have surged due to repeated attempts. Developers have observed that the model is no longer capable of managing complex engineering tasks, displaying problems such as performance regression and an inability to activate appropriate planning modes. Regarding the root cause of the intelligence decline, some hypothesize that it may stem from high computational demands and costs, with Anthropic potentially adjusting the default thinking level to strike a balance between intelligence, latency, and costs. Furthermore, there are conjectures that Anthropic might be selectively degrading intelligence based on IP, which essentially distills down to an issue of computational resource allocation.