OpenAI cuts GPT-5.6 Luna price 80%

OpenAI's price cuts are tied to a broader argument about what makes AI infrastructure valuable. The company emphasizes that value comes not from scale alone but from what intelligence enables: more capable models, available to more people, at lower cost. This is the logic behind its recent engineering work, where GPT-5.6 Sol helped optimize the production serving software and improve token-generation efficiency, cutting costs and increasing output.

These efficiency gains feed into a broader feedback cycle OpenAI describes, where better models drive adoption, which supports further investment. The company points to recent benchmark results where improvements to retained reasoning and context management raised GPT-5.6 Sol's score on the public ARC-AGI-3 task set from 13.3% to 38.3% while using six times fewer output tokens. The model itself did not change, only the surrounding system, underscoring OpenAI's "full stack" approach.

The stakes, as OpenAI frames it, hinge on whether lower prices truly expand what becomes practical for customers. The company argues that a stronger, more efficient model can ultimately be more economical than a cheaper model requiring repeated attempts. The risk is whether the cost-per-successful-outcome math holds across diverse workflows, which will likely determine whether this pricing strategy broadens adoption or simply compresses margins.