Updated
Updated · Ars Technica · Aug 14
OpenAI, Anthropic Cut AI Model Prices by Up to 80% as Chinese Rivals Gain Ground
Updated
Updated · Ars Technica · Aug 14

OpenAI, Anthropic Cut AI Model Prices by Up to 80% as Chinese Rivals Gain Ground

3 articles · Updated · Ars Technica · Aug 14

Summary

  • OpenAI cut GPT-5.6 Luna prices by 80%, while Anthropic launched Claude Opus 5 at half the price of its top-end Fable 5, escalating a discount push among leading US AI labs.
  • Almost 25% has been wiped off prices paid for top US models since mid-July, according to Silicon Data’s token price index, as companies rein in AI spending and hunt for cheaper options.
  • DoorDash and Airbnb are already using Chinese-made models to curb bills, reflecting how developers such as Moonshot and DeepSeek have expanded from Silicon Valley to Europe.
  • The price war marks a shift for US groups that long competed mainly on performance, as increasingly capable open Chinese models pressure closed-model providers ahead of planned IPOs at trillion-dollar valuations.

Insights

Will OpenAI’s 14x faster GPT-5.6 Sol revolutionize real-time incident response, or simply accelerate the blast radius of AI errors?
How does the new disaggregated hardware architecture achieve 750 tokens per second without sacrificing OpenAI's frontier-level intelligence?
With AI now operating at ultra-low latency, can human oversight possibly keep pace before a critical financial or security misstep occurs?