OpenAI’s 80% Luna Cut Turns AI Pricing Into a New Battlefield

OpenAI’s 80% Luna Cut Turns AI Pricing Into a New Battlefield
OpenAI’s steep price reduction for its GPT-5.6 Luna model is being presented as an efficiency breakthrough. For customers scrutinizing escalating AI bills, it is also a sign that the economics of frontier models are becoming harder to defend.

The company cut Luna’s API price by 80% to $0.20 per million input tokens and $1.20 per million output tokens, while lowering the mid-range Terra model by 20%, to $2 and $12 respectively. Its most capable model, Sol, did not receive a standard price cut, though OpenAI introduced a Fast mode that it says can run up to 2.5 times faster at twice the price.

OpenAI frames the change as the result of a full-stack effort: improving models, GPU-serving software, routing and the agent systems that manage tools and context. It says Sol-assisted production optimizations reduced end-to-end serving costs by 20% and lifted token-generation efficiency by more than 15%. The company’s broader argument is that buyers should judge AI by the cost of a completed task—not the nominal price of tokens—because a stronger model can be cheaper overall if it avoids retries and human oversight.

That rationale lands amid a more skeptical market. EMARKETER analyst Jacob Bourne said “the era of tokenmaxxing is over,” arguing that enterprises are pushing back after finding how easily usage can rise without commensurate value. Gartner’s Arun Chandrasekaran, meanwhile, described the cuts as evidence that frontier vendors are shifting away from rigid commercial terms toward more flexible pricing. Reporting also points to pressure from cheaper Chinese open-weight models and competition with Anthropic as customers focus more heavily on price.

Greg Brockman called Luna “by far the most price-efficient model in its class,” while saying the company’s research is aimed at making models efficient at each intelligence level. Sam Altman distilled the strategy more cautiously: OpenAI wants “the best price/intelligence tradeoff at every level.”

Continue reading https://foxvector.com/stories/019fca8b-4758-276e-7125-1c2d65a3320a