Anthropic's Sonnet 5.5 model, codenamed "Fennec," is likely to be released next month, supporting a maximum context window of 2 million tokens, up from Sonnet 5's 1 million. The leak indicates faster reasoning, lower latency, improved long-context reasoning and multi-step planning, and enhanced tool-calling abilities for browsers and terminals. Its performance approaches the flagship Claude Fable 5, but pricing will remain in the Sonnet tier, targeting cost-effectiveness. The model's development is in its final stage, directly targeting DeepSeek V4 Flash. The competition among large models has shifted to cost-effectiveness, focusing on intelligence per yuan spent. DeepSeek V4 Flash disrupted the market with ultra-low pricing, pressuring overseas companies. Anthropic's Sonnet 5.5 is a direct response, offering near-Fable 5 capabilities at Sonnet prices. Anthropic's Haiku series has not been updated for about a year, and analysts suggest Sonnet may take over its mid-range market position. This strategy could compress DeepSeek's price advantage and avoid internal competition, but DeepSeek has established a low-price mindset, and OpenAI's Luna is capturing lightweight scenarios. Whether Sonnet 5.5 can claim the cost-effectiveness crown will be seen at its official release next month.