US AI Labs Slash Model Prices in Token Cost War With Chinese Rivals
Leading developers cut API prices as corporate clients switch to cheaper models
Leading American AI developers, including OpenAI and Anthropic, are slashing token pricing for mid-tier AI models as cost-conscious enterprise clients increasingly turn to cheaper alternatives from Chinese rivals. The price cuts mark a major tactical shift for U.S. frontier labs that previously focused on raw performance rather than pricing competition.
Key details
Rising computational costs have forced corporate clients to restrict AI API usage and explore lower-cost options, driving demand toward open and cut-price models developed by Chinese firms like Moonshot AI and DeepSeek. Companies such as DoorDash and Airbnb have reportedly integrated Chinese-developed models to curb mounting software and infrastructure bills.
In response to the competitive shift, OpenAI reduced the API pricing for GPT-5.6 Luna by 80 percent, cutting input token costs from $1.00 to $0.20 per million tokens and output token costs from $6.00 to $1.20 per million tokens. Similarly, Anthropic released Claude Opus 5 at $5.00 per million input tokens and $25.00 per million output tokens—half the pricing of its flagship Fable 5 model—while canceling a scheduled price increase for its Sonnet 5 model. According to Silicon Data's token price index, these pricing adjustments have driven down overall customer expenditure on U.S. frontier models by nearly 25 percent since mid-July 2026.
Why this matters
The token price war highlights a growing economic reality in artificial intelligence: infrastructure costs and compute consumption are dictating enterprise adoption. As U.S. providers transition corporate customers away from flat-rate subscriptions toward usage-based billing tied to compute consumption, companies are actively managing token budgets.
Context
The price pressure comes at a critical juncture for U.S. AI developers. Both OpenAI and Anthropic are planning initial public offerings targeting trillion-dollar valuations, requiring them to demonstrate sustainable revenue growth amid immense capital expenditure. As developers invest tens of billions into data center infrastructure and GPU procurement, price compression on mid-tier model tokens could impact profit margins across the industry.
What happens next
Enterprise customers are expected to continue benchmarking open-source and foreign models against proprietary U.S. API offerings to optimize operational budgets. U.S. labs will likely accelerate efficiency improvements in inference hardware and software architectures to maintain competitive margins while accommodating lower token prices.
Source: Ars Technica Published on AI Usage Global, author: AUG Bot
