OpenAI and Anthropic Slash AI Prices as Chinese Rivals Close the Gap

OpenAI has reduced the price of its lightweight GPT-5.6 Luna model by 80%, bringing API costs to $0.20 per million input tokens and $1.20 per million output tokens, and has trimmed its mid-tier Terra model by around a fifth. Anthropic has responded with Claude Opus 5, which the company says delivers frontier-level performance at roughly half the cost of its previous flagship.

The driving force behind both moves is competition from Chinese AI labs. DeepSeek, Moonshot's Kimi and Zhipu's GLM have closed the capability gap enough that enterprise buyers are finding it hard to justify the premium for Western models. Running a workload through Zhipu's GLM has been benchmarked at roughly $544, compared with $4,811 for the same job on Claude — a near ninefold difference that finance teams are no longer willing to overlook.

OpenAI CEO Sam Altman acknowledged the pressure directly on social media, stating that GPT-5.6 Sol already undercuts Anthropic's Fable 5 and that OpenAI would be willing to deliver at one-quarter of the price if competition demands it.

The price war carries consequences beyond cheaper API bills. Geopolitically, Anthropic reportedly had just a 90-minute window to withdraw Fable 5 access for foreign nationals, and OpenAI's top GPT-5.6 tier has been restricted to roughly 20 approved partners. DeepSeek, meanwhile, is reportedly targeting a $74 billion valuation ahead of a China listing, raising live questions about which sovereign and enterprise buyers it can realistically serve outside China. Infrastructure analysts warn that cheaper tokens are likely to drive higher inference volumes, directing more workloads toward the power-dense facilities that already command 60% lease rate premiums over conventional data centres.