Recently, Anthropic unveiled its cost-effective and lightweight model, Claude Haiku 5.5. This new offering comes at a mere one-tenth of the price of its forerunner, with the base rate for short-context API even undercutting that of the domestic compact model, DeepSeek-V4.1-Flash. Concurrently, OpenAI declared the complete launch of GPT-6 and made its latest compact model, GPT-6 Luna, accessible at no cost within the free tier of ChatGPT. Both of these international AI powerhouses have concurrently slashed prices for their compact models. However, these firms are not luring customers with unsustainable, below-cost pricing. Instead, they are employing a multi-tiered pricing approach across their flagship, mainstay, compact models, and free offerings. By leveraging cost-effective compact models to boost user engagement, they aim to generate profits through subsequent high-value services, including complex task execution at a premium, enterprise solutions, and developer subscriptions.
This round of price cuts has a relatively limited impact on the domestic market. Leading domestic models have established barriers through their superior Chinese language processing capabilities, tailored local services, and the demand for private deployments. This makes it challenging to disrupt the existing market dynamics in the near future. Nevertheless, they will encounter some competitive pressures in overseas markets. Domestic vendors should avoid getting embroiled in reckless price wars. The crux lies in refining their own multi-tiered business frameworks and constructing a diversified revenue model that does not hinge solely on differences in token pricing.