OpenAI and Anthropic cut prices to hold on to customers

Prices that customers pay at the leading US AI labs fell by almost a quarter from mid-July to mid-August 2026. That is according to Silicon Data's index of token prices, which the Financial Times has reported on. A token is a small chunk of text that the AI model works with, and it is the unit in which pricing is typically set.

According to the report, OpenAI and Anthropic have launched cheaper models to hold on to customers who are moving to cheaper Chinese alternatives. Rising AI bills are prompting companies to limit usage and look for cheaper models. That is helping Chinese developers such as Moonshot and DeepSeek gain a foothold with users from Silicon Valley to Europe. The Financial Times assesses that Chinese models can increasingly match the leading American ones, but at lower cost.

Big cuts at OpenAI

On July 30, OpenAI cut the price of GPT-5.6 Luna by 80 percent, three weeks after its launch. The price went from $1 to $0.20 per million input tokens and from $6 to $1.20 per million output tokens. The mid-tier model GPT-5.6 Terra became 20 percent cheaper at the same time. The price of Sol is unchanged, but a fast mode has been introduced at double the price and 2.5 times the speed.

OpenAI attributes the cuts to efficiency gains: 20 percent lower total operating costs and more than 15 percent better efficiency in token generation. Those are the company's own figures and have not been independently verified. Anthropic has launched Claude Opus 5, which is described as frontier intelligence at half the price of the company's most capable model, Fable 5.

Chinese price competition

In a per-task price comparison, DeepSeek V4 Pro costs around 4 cents, Moonshot's Kimi K3 costs 94 cents and OpenAI's Sol costs $1.04. In April, DeepSeek cut the input price of V4-Pro by 75 percent, and according to The Standard, usage rose almost fourfold after the cut. However, on August 13, DeepSeek announced a new peak-hour price of four times the current one, which is still expected to remain below its competitors'.

OpenAI itself says the company is building a robust infrastructure portfolio and matching each workload with the systems that suit it best.

For businesses and users, this could mean cheaper access to AI. However, the comparisons use different units and are not directly comparable.