2026年6月至9月间，主要人工智能开发商OpenAI和Anthropic对其模型阵容（包括GPT-6 Sol、GPT-5.6 Luna和Claude Opus 5.5）实施了高达80%的激进降价。
OpenAI and Anthropic Slash LLM Pricing by Up to 80% Amid Open-Source Pressure and Looming IPOs Between June and September 2026, major artificial intelligence developers OpenAI and Anthropic enacted aggressive price cuts of up to 80% across their model lineups, including GPT-6 Sol, GPT-5.6 Luna, and Claude Opus 5.5.
此次降价显著影响了缓存读取成本，而缓存读取成本占智能体编程支出的大头。
The reductions notably impacted cache read costs, which represent the bulk of agentic coding expenditures.
这场价格战主要由日益加剧的商品化、推理算力效率提升，以及来自DeepSeek和Moonshot AI等开放权重模型提供商的激烈竞争所驱动，后者的模型广泛托管在OpenRouter等第三方推理平台上。
The price war is primarily driven by escalating commoditization, improved inference compute efficiency, and intense competition from open-weight model providers such as DeepSeek and Moonshot AI, whose models are widely hosted across third-party inference platforms like OpenRouter.
这些大幅降价发生在潜在首次公开募股（IPO）之前，引发了关于在数十亿美元研发支出下推理利润率和收入可持续性的疑问。
These steep cuts come ahead of prospective initial public offerings, raising questions about inference profit margins and revenue sustainability amid billions in research spending.