OpenAI and Anthropic cut model prices due to competition from China

OpenAI says it has cut the price of GPT-5.6 Luna, which it describes as its “fastest and most affordable model”, by 80 percent. Anthropic, meanwhile, has released Claude Opus 5, describing it as a system with “peak intelligence” at half the price of Fable 5, the most powerful model from Anthropic to date.

According to the token price index from Silicon Data, these moves have helped bring down the prices customers pay for models from leading US labs by almost a quarter since mid-July. Tokens are units of data processed by language models and are used to calculate a large share of customer bills.

The price cuts mark a shift for US companies producing closed models, which have so far competed primarily on performance. The growing capabilities of open Chinese models, which developers can freely download and modify, are contributing to pressure on prices. The price cuts also come as OpenAI and Anthropic plan to go public at valuations in the trillions of dollars, and investors demand evidence that substantial investments in AI are delivering returns.

Corporate customers are also burdened by the shift among some businesses from flat-rate subscriptions to billing based on actual consumption of computing resources, which Anthropic and OpenAI are introducing. Some companies are therefore imposing limits on AI usage or testing cheaper alternatives — DoorDash and Airbnb said they had started using models made in China to save costs. At the same time, a wave of new models from Chinese labs is narrowing the performance gap with leading US models, raising concerns in the US technology industry about losing customers despite high spending to maintain a technological lead.