OpenAI Rolls Out GPT-6 Sol and Luna at Half the API Price

OpenAI has extended its GPT-6 generation with new versions of its smaller Sol and Luna models, TechCrunch reported on September 22, 2026. The headline change for developers is price: the new models cost half as much through the API as the 5.6-series versions of Sol and Luna.

Earlier in the month, OpenAI had launched GPT-6 Astra as its most capable model. The company described the new releases as a way to make that generation’s intelligence “more efficient and accessible.”

Two tiers

Sol and Luna sit at different levels of OpenAI’s lineup. According to TechCrunch, Sol is designed for complex tasks such as coding, while Luna is aimed at “high-volume tasks with a clear goal, like summarizing documents, extracting information, or answering quick questions.”

OpenAI attributed the price cut to improvements in caching and inference, the process of running a trained model to answer requests. For businesses that send large volumes of routine requests, halving the per-request cost can change which tasks are worth automating.

Fewer mistakes, by OpenAI’s measure

OpenAI also claimed better accuracy. “On our internal factuality evaluation, which is based on de-identified real-world conversations where users flagged mistakes by our models, GPT-6 Sol makes about half as many mistakes as its predecessor,” the company said, according to TechCrunch.

That claim rests on OpenAI’s own internal test, not an independent benchmark, so it is a company figure rather than a third-party result. TechCrunch also noted that, as usual, OpenAI said the new models outperform its main rival’s top models, and that the rival released an update 90 minutes before OpenAI’s announcement, a sign of how close the competition has become.

Who gets it

The new Sol and Luna are available to most paid accounts in ChatGPT Work and Codex, and in the API. Free and Go users get Luna. TechCrunch reported that OpenAI planned to roll the models out gradually to the ChatGPT app and website over the course of the day.

What it means in practice

For US developers, the main effect is a lower bill for the same class of work. Teams already building on the 5.6-series models can test whether the new versions match their quality needs at the lower price. The factuality claim is worth checking against their own data, since an internal evaluation built from flagged ChatGPT conversations may not reflect every use case.

The release also landed in a crowded week for model launches. A day earlier, xAI shipped Grok 4.7 at a low per-token price; see our report on Grok 4.7. The price competition among US labs is now as visible as the race on capability.