OpenAI and Anthropic in price war as Chinese AI rivals gain ground

Details

The US AI price war intensified this week with two major moves: OpenAI cut GPT-5.6 Luna input token prices from $1 to $0.20 per million (80% reduction; output dropped from $6 to $1.20), and Anthropic launched Claude Opus 5 at $5/$25 per million input/output tokens — half the price of its flagship Fable 5. According to Silicon Data's token price index, customers are paying nearly 25% less for leading US lab models than they did in mid-July. The cuts are a direct response to Chinese competition. Moonshot, DeepSeek, and other open-weight Chinese models have made enough inroads that DoorDash and Airbnb are now publicly using them to rein in AI costs. Chinese models are freely downloadable and customizable, creating structural pricing pressure on proprietary US offerings that is not easily reversed. A secondary driver: OpenAI and Anthropic are shifting enterprise customers from flat subscriptions to usage-based billing, causing some companies to hit unexpected cost spikes and respond by capping AI usage or testing Chinese alternatives. Both companies are preparing IPOs at trillion-dollar valuations while needing to demonstrate positive unit economics. For practitioners: the most aggressive cuts are landing in the mid-tier (not just cheapest models), making it worth reassessing current model selection and cost models. The 80% Luna price cut applies to one of OpenAI's fastest, most affordable models.