OpenAI Just Slashed Its New Model Prices in Half

OpenAI dropped two new models this week, Sol and Luna, and buried in the announcement is the part that actually matters to anyone paying an API bill: prices are cut in half compared to the previous generation.

Sol is built for the heavier lifting, coding, multi-step tasks, the stuff where you need the model to actually think. Luna is the workhorse for everything else, summarizing a document, pulling a number out of a report, answering a quick question fast. OpenAI trained both using the same approach behind its flagship Astra model, then squeezed it down into something smaller and cheaper to run.

And the mistake rate dropped too. OpenAI says Sol makes about half as many errors as its predecessor did. That is the detail I actually care about. Cheaper is nice. Cheaper and more reliable at the same time means a small team can afford to automate a lot more than it could last year.

A year ago, running a capable model on every document a company touched was a real cost decision, something you had to budget for and justify. At half the price and with fewer errors, that decision gets a lot easier to make. More teams will just do it.

There is a quiet pattern behind the flashy launches. Every few months the top tier gets smarter, and the tier below it gets cheap enough that everyone else can use it too. Luna costs ten cents per million input tokens now. That is not a typo, and it is a sign of where all of this is heading.

Good tools getting cheaper and more available is the story I want more of. Not everyone needs the biggest model. Most people just need one they can afford to run all day.

——

Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277

——

Sources:
https://openai.com/index/introducing-gpt-6-sol-and-luna/
https://techcrunch.com/2026/09/22/openai-launches-gpt-6-sol-and-luna/
https://www.ghacks.net/2026/09/23/openai-launches-gpt-6-sol-and-luna-with-50-lower-api-pricing-than-gpt-5-6/

Repost this. Thanks.