GPT-5.6 Sol pricing cut 50%

What's this about?

GPT-5.6 Sol's price dropped by half. HN's most important contribution was a correction: this is the price on OpenRouter, not OpenAI's native API pricing, which is unchanged. The thread's substantive question is what a cut this size reveals — if Sol can halve and Luna can drop 80% while both continue being served, the original prices carried very large margins, and both may still be profitable.

What exactly is a price war in inference?

Serving a model has a real cost — GPUs, power, engineers — but the price charged is a business decision that can sit far above it. When several providers reach similar capability, price becomes the differentiator, and each cut pressures the others. What makes this specific to AI is switching cost: because routers and OpenAI-compatible APIs make swapping providers nearly free, there's little to hold customers in place. The winner is whoever can serve most cheaply, which is why efficiency work and custom silicon matter more than they appear to.

What's the community saying?

1. The OpenRouter caveat was the top correction. The headline reads like an OpenAI price cut and isn't one — worth knowing before you plan around it.
2. "If they can cut Sol 50% and Luna 80%, the original price carried a massive margin." The thread's central inference, with the honest addendum that we'll never actually know.
3. The precedent worked. Luna's usage jumped substantially after its own cut, making it one of the more competitive models on OpenRouter — so this looks like OpenAI testing how much share Sol can take.
4. "Is this motivated by the value of the thinking traces?" A recurring suspicion — cheap inference as a way to acquire reasoning traffic, which echoes the token-broker distillation theory from yesterday.
5. A complaint about the model itself. Several find 5.6 worse at simple tasks and prone to over-complicating: asked for a to-do list, it returned a four-page essay, where 5.4 returned checkboxes.
6. Scale check. One user reported burning over a billion tokens a day of Sol at xhigh effort for five days before hitting a reset limit, calling the $200 plan still the best deal available.