Recently, the global large model market has seen a significant adjustment in prices and ecosystems. Anthropic officially launched the low-cost, lightweight model Claude Haiku5.5, with its unit price dropping directly to one-tenth of the previous generation’s price. The bare API price for short context even falls below that of some domestic small models. At the same time, OpenAI also announced the full launch of GPT-6, and it will make the same-generation small model GPT-6Luna available at the free tier of ChatGPT. These two major overseas giants simultaneously reduced the prices of small models during this period, drawing great attention from the industry.
From a deep business logic perspective, these two companies do not rely on low prices to incur losses in order to attract users maliciously. Instead, they have established a mature “tiered pricing system”. They use low-price small models as a traffic entry point to attract a large number of users, while true profits come from high-price complex task processing, enterprise-level services, and developer subscriptions—such higher-level services. This scientific business pyramid structure ensures the sustainability of the low-price strategy.
For the domestic large-scale model market, this wave of low prices overseas currently has a relatively limited impact. Due to its advantages in the Chinese language, high-quality services tailored to local needs, and widespread private deployment requirements, domestic leading models are unlikely to be affected by overseas products in the short term. However, in the overseas market, domestic manufacturers will inevitably face some pressure from user diversion.
Faced with this trend, domestic manufacturers need not blindly follow the homogenized price war in overseas markets. The current core task is to improve and perfect their business structure, and build a diversified revenue system that does not solely rely on token price differences, so as to stand firm and remain unbeatable in the intense global competition.