The Financial Times reported on Friday (14) that consumers are paying almost 25% less for artificial intelligence models from leading US laboratories than in mid-July, after price cuts announced by OpenAI and Anthropic, according to data from Silicon Data.
OpenAI cut the price of the intermediate model GPT-5.6 Luna by 80%, while Anthropic launched Claude Opus 5 at half the price of Fable 5. The cuts come at a time when companies are seeking to contain costs after the rise in AI bills and when Chinese developers such as Moonshot and DeepSeek gain ground among American and European customers with cheaper models that are increasingly closer to the capabilities of US leaders.
OpenAI announces cuts in July
When announcing cuts to selected models on July 30, OpenAI said the changes were intended to improve the performance of models per dollar on enterprise workloads. At the time, the company cut the price of GPT-5.6 Luna by 80%, reduced that of GPT-5.6 Terra by 20%, and accelerated the performance of GPT-5.6 Sol in the API, without changing the price of the frontier model.
“We are building a resilient infrastructure portfolio and combining each workload with the most appropriate systems to run it,” OpenAI said. “This approach supports the two extremes of the price-performance curve.”
The laboratories maintain or increase prices for top-tier models, with the justification that a more capable and more expensive model can complete a task with fewer tokens and therefore at a lower cost. DeepSeek will add peak prices for its V4 models, which will quadruple current levels, although the new values remain below the main competitors.
