Leading US AI labs OpenAI and Anthropic are cutting prices for their models as they face growing competition from Chinese developers like Moonshot and DeepSeek. The price war reflects rising AI costs that are forcing companies to seek cheaper alternatives. OpenAI recently reduced prices for its GPT-5.6 Luna model by 80%, while Anthropic launched Claude Opus 5 at half the cost of its flagship Fable 5.

According to Silicon Data’s token price index, customer costs for US lab models have dropped nearly 25% since mid-July. These cuts mark a shift for US AI companies that previously focused on performance over affordability. Chinese open-source models, which can be freely downloaded and modified, have gained traction among users in Silicon Valley and Europe.

Companies such as DoorDash and Airbnb have begun testing Chinese-made models to control expenses. The trend coincides with OpenAI and Anthropic preparing for initial public offerings at high valuations, increasing pressure to demonstrate profitability. Both companies are also moving enterprise clients from flat subscriptions to usage-based billing, contributing to higher bills.

While headline token prices offer limited comparison, more efficient models may complete tasks faster despite higher per-token costs. Artificial Analysis found Anthropic’s Opus 5 at medium effort matched Moonshot’s Kimi K3 at max effort in cost and performance.