The price war in AI has broken out in earnest. According to the Financial Times, OpenAI and Anthropic are slashing prices for some of their models as Chinese rivals such as DeepSeek and Moonshot squeeze the market with cheaper, increasingly capable alternatives.
OpenAI has cut the price of GPT-5.6 Luna — billed as its "fastest and most affordable model" — by 80 percent, from $1 to $0.20 per million input tokens, and from $6 to $1.20 per million output tokens. A second tier, Terra, was reduced by about 20 percent. Anthropic, for its part, has positioned its newly launched Claude Opus 5 at $5 per million input tokens and $25 per million output tokens — roughly half the price of its top-end Fable 5 model — and cancelled a planned September price increase for Sonnet 5.
The moves reflect two pressures at once. Enterprises, after two years of experimentation, are moving AI workloads into production and watching every dollar; buyers increasingly compare total cost per completed task rather than sticker price per token. At the same time, Chinese models are being seriously evaluated by Western companies — DoorDash and Airbnb are among those reported to have tested them — giving procurement teams leverage in negotiations even when the US models win on quality.
The FT frames the moment as a turning point: after years of competing mostly on capability, American labs now face the economics of the market. The open question is whether steep cuts can be sustained while funding the enormous compute costs behind frontier development — and whether the AI race is becoming, at least in part, a mature market where price matters as much as performance.
