OpenAI and Anthropic slash prices as Chinese AI rivals win over cost-conscious customers
A price war is breaking out at the top of the AI industry. American labs are cutting what they charge for their models, and the trigger is clear: cheaper tools from China are pulling customers away.
Key points
- OpenAI cut prices for its GPT-5.6 Luna model by 80 percent, describing it as the company's "fastest and most affordable" option.
- Anthropic launched Claude Opus 5 at half the price of its most capable model, Claude Opus 5 Fable.
- Chinese AI developers including Moonshot and DeepSeek are winning users in Silicon Valley and Europe by undercutting US rivals on price.
- Rising AI costs are pushing companies of all sizes to hunt for cheaper models or simply use AI less.
Something changed quietly over the last year. Paying for AI stopped feeling free.
Businesses that built their workflows around tools like OpenAI's GPT models began getting larger bills. Some cut back. Others started shopping around. And a handful of Chinese AI developers were ready with models that cost considerably less.
What exactly happened?
Two of America's most prominent AI labs responded to the pressure this week with significant price cuts. As first reported by Ars Technica AI, OpenAI said it was reducing prices for GPT-5.6 Luna, its entry described as the company's "fastest and most affordable model", by 80 percent. Separately, Anthropic launched Claude Opus 5, a large language model (the technology behind chatbots like ChatGPT and Claude) that the company says delivers "frontier intelligence at half the price" of Fable 5, its most powerful and expensive option.
These are not small discounts. An 80 percent cut means a bill that was $100 drops to $20 for the same work.
Who is pushing them to do this?
The competitive pressure is coming from China. Developers including Moonshot and DeepSeek have been releasing capable AI models at prices that undercut Western rivals. Customers from startups in San Francisco to businesses in Europe noticed. Some switched.
DeepSeek attracted particular attention earlier this year when it released models that matched top US performance at a fraction of the running cost. That sent a clear signal to the industry: price now matters as much as quality.
What does this mean for ordinary users and businesses?
For anyone paying to use AI tools, the timing is good. Competition between labs tends to push prices down across the board, and that benefit eventually reaches the products built on top of these models, including the apps employees use at work and the services customers interact with daily.
The risk, which is worth watching, is whether cheaper models cut corners. A lower-priced model is only good value if it still does the job reliably. Businesses evaluating any new AI tool, whether from a US lab or a Chinese developer, should test it on real tasks before committing.
For now, the price war looks set to continue. Neither OpenAI nor Anthropic has suggested these cuts are the last.
Common questions
Does a cheaper AI model mean a worse one?
Not necessarily. Labs often reduce prices on older or narrower models while keeping their most powerful options at a premium. The key is matching the model to the task: a cheaper model may handle most everyday jobs perfectly well.
Should businesses be concerned about using AI tools from Chinese developers?
It depends on what data those tools process. If sensitive customer or company data is involved, businesses should check where that data is stored and who can access it before choosing any provider, regardless of where it is based.
