OpenAI and Anthropic are moving into a more aggressive pricing phase as businesses look harder at the cost of using AI. The pressure is coming from two directions at once: rising bills for corporate users and cheaper Chinese models that are becoming harder for customers to ignore.
The result is a sharper contest over mid-tier AI models, where price cuts can help US labs keep users who may otherwise test alternatives from Moonshot, DeepSeek, and other Chinese developers.
Why AI model prices are under pressure
For much of the recent competition among leading US AI labs, the central selling point has been performance. Proprietary, closed models from companies such as OpenAI and Anthropic have been positioned around capability, especially for customers seeking the strongest systems available.
That balance is changing. The source article describes a market where companies are increasingly sensitive to AI spending, especially as some enterprise customers move away from flat subscriptions and toward usage-based billing. Under that model, businesses pay based on the computational resources they consume.
This matters because AI usage can grow quickly inside an organization. If more teams use large language models for coding, analysis, customer support, research, or internal tools, the bill can become a live operating concern rather than an experimental expense.
Some businesses have responded by setting caps on AI usage or testing cheaper systems. DoorDash and Airbnb are named in the source as companies that have started using Chinese-made models to help control bills.
The price cuts from OpenAI and Anthropic
OpenAI recently said it was cutting prices for GPT-5.6 Luna, described by the company as its “fastest and most affordable model”, by 80 percent. The change took the model from $1 to $0.20 per million input tokens and from $6 to $1.20 per million output tokens.
Anthropic, meanwhile, launched Claude Opus 5 and presented it as offering “frontier intelligence… at half the price” of Fable 5, which the source identifies as the company’s most capable model. Opus 5 is priced at $5 per million input tokens and $25 per million output tokens.
The source also says Anthropic called off a planned price increase for Sonnet 5 that had been due to take effect from September.
According to Silicon Data’s token price index, the recent moves have helped reduce what customers pay for models from leading US labs by almost a quarter since mid-July. That figure points to a broader pricing reset, not only a single promotional discount.
Why token prices do not tell the whole story
AI model pricing is usually described in tokens. Tokens are the units of data processed by language models, and they are used to calculate many customer bills. Input tokens measure what is fed into the model, while output tokens measure what the model generates in response.
Still, headline token prices can be misleading. A model with a higher listed price may solve a task with fewer tokens or fewer attempts. In that case, it can end up costing less for the completed job than a cheaper-looking option.
There is another variable: effort settings. Many models can run at different effort levels, which change the computing power used to answer a question. Those settings can affect both performance and the final cost of a task.
Artificial Analysis, which benchmarks models in areas including math, science, coding, and reasoning, found that Anthropic’s Opus 5 at “medium” effort delivered similar performance and cost per task to Moonshot’s Kimi K3 at “max” effort. It also found that OpenAI’s GPT-5.6 Luna at “max” effort performed similarly to DeepSeek’s V4 Flash at “max,” while costing just under twice as much per task.
Chinese AI rivals are changing customer behavior
The price competition is being shaped by Chinese models that are both cheaper and increasingly capable. The source says Chinese developers including Moonshot and DeepSeek are gaining ground with users from Silicon Valley to Europe.
These models also differ in how developers can use them. The article describes increasingly capable “open” Chinese models that can be freely downloaded and modified by developers. That contrasts with the proprietary “closed” models sold by leading US AI groups.
For customers, the practical question is no longer simply which model scores highest in a general benchmark. The more immediate question is which model can complete the required work at an acceptable quality and cost.
That creates pressure in the middle of the market. The latest US price cuts apply to mid-tier products, making them more competitive with Chinese offerings while leaving the very top of the market more protected.
The bigger stakes for US AI labs
The timing is important because OpenAI and Anthropic are also planning initial public offerings at trillion-dollar valuations, according to the source. Investors are looking for evidence that the industry’s heavy AI spending can produce returns.
Lower prices can help keep customers from leaving, but they also raise harder questions about margins and long-term business models. If customers become trained to compare models mainly by task cost, even the most recognized AI labs may have to defend their pricing more often.
OpenAI and Anthropic declined to comment. A person close to Anthropic said Opus 5’s pricing below Fable 5 was how the startup’s “family of models is built, so there’s no connection to competitors.”
Mantas Lukauskas, AI tech lead at Hostinger, said prices for the strongest models were “flat to rising.” He described the recent pricing changes as the “first real test” of whether groups such as Anthropic and OpenAI can protect the cost of their most advanced systems: “The US labs have cut the middle and are defending the top.”
That framing captures the current market well. The contest is not only about cheaper AI. It is about whether US labs can keep premium pricing for their most capable models while cutting enough in the middle to stop customers from moving elsewhere.