OpenAI and Anthropic, the two most prominent AI companies in the West, are both cutting prices sharply within weeks of each other. OpenAI cut the price of GPT-5.6 Luna, its fastest and cheapest model, by 80 percent. Anthropic followed with Claude Opus 5, a model it prices at half of what it charged for Fable 5, its previous top tier model.
This is not normal competitive jockeying between two companies chasing the same customers. It is a defensive move against a third party. Both companies are losing ground to Chinese AI makers, especially DeepSeek and Moonshot, whose models cost a small fraction of what OpenAI and Anthropic charge for comparable quality.
The gap in price is not small. DeepSeek's newest fast model charges close to $0.14 for every million words worth of text it reads, and $0.28 for every million words it writes. Anthropic's new Opus 5 starts at $5 per million words in and $25 per million words out. Depending on how it is used, that puts the Chinese model at roughly a fraction of the cost.
This price gap is already showing up in real business decisions, not just spreadsheets. DoorDash has said publicly it uses a Moonshot model because it found better quality at a lower price. Airbnb and the German industrial group Siemens have also been reported testing Chinese AI models to control their own AI spending. Cursor, a coding tool being bought by SpaceX for around 60 billion dollars, built one of its products on top of a Moonshot model as well.
The scale of the shift is bigger than a handful of company names. Tracking data from OpenRouter, a marketplace that routes AI traffic to different providers, showed Chinese made models capturing as much as 46 percent of business AI usage in a single week this year. A year earlier, that figure was close to nothing.
There is a real cost to going cheap, though. US lawmakers have opened an investigation into companies, including DoorDash, over using AI models built in China, citing concerns about where business data ends up and who controls the technology behind it. For any company running customer data, financial numbers, or internal documents through an AI model, the country where that model was built now matters almost as much as the price on the invoice.
For most businesses, this price war is good news in the short run. AI features built into the software you already use, things like customer service bots, spreadsheet helpers, or coding assistants, will likely keep getting cheaper as OpenAI and Anthropic defend their position. But if you or your vendor is picking a Chinese model directly to save money, treat it like any other outsourcing decision. Check where your data goes, not just what you pay for it.