What's this about?
OpenAI dropped the price of GPT-5.6 Luna — its fastest, cheapest model — by 80% (roughly 5× cheaper), framing it as advancing the "price-performance frontier." Part of the savings comes from the model optimizing OpenAI's own serving stack, cutting end-to-end serving cost ~20% and improving token-generation efficiency 15%+.
What exactly is "price-performance frontier"?
For any task there's a trade-off between how good a model is and how much it costs to run. The "frontier" is the best deal available at each quality level — the most capability per dollar. Pushing the frontier means giving you the same quality for less money (or more quality for the same money). An 80% cut on an already-cheap model shoves that frontier down hard.
What's the community saying?
HN 506 pts, 336 comments
- "The Chinese models are really making a dent" — many read the cut as a direct response to cheap, capable open models like Kimi K3.
- Making an already-cheap model 5× cheaper is "crazy"; people who use Luna at home and Sol at work are re-evaluating.
- Note that OpenAI didn't mention Grok at all in the framing.
- Some unease at model *segmentation* — pushing users to pick exactly which tier solves their problem.