OpenAI cuts GPT-6 Sol and Luna API prices in half

What shipped

GPT-6 Astra opened what OpenAI calls a new generation of models. Sol and Luna are meant to make that generation more efficient and accessible.

The first Sol and Luna models arrived at the beginning of this year and occupied different positions in OpenAI's model hierarchy:

The biggest change is the API price. OpenAI says GPT-6 Sol and Luna cost half as much as Sol and Luna from the 5.6 series. The company attributes the reduction to improvements in caching and inference.

OpenAI also says the updated models produce more factually accurate answers and make fewer mistakes on programming tasks.

The error claim

OpenAI's internal factual-accuracy evaluation uses anonymized real-world conversations in which users had flagged model errors. According to the company, GPT-6 Sol makes about half as many errors as its predecessor and reaches Astra-level reliability at a substantially lower cost.

That comparison is the more consequential part of the release. A cheaper model with similar reliability can change which tasks developers are willing to run continuously, while a smaller error rate matters more than another broad claim about general capability.

Still, the announcement is quiet about the measurement itself. It provides a relative improvement, not a raw error rate or a detailed breakdown of where Sol improves most. What I'd want to know is whether the cost reduction survives the hardest workloads, rather than only the average one.

A release shaped by competition

As it usually does after a model launch, OpenAI says its new products outperform Anthropic's leading models, including Fable and Opus. Anthropic released a new version of Opus 5.5 just 90 minutes before OpenAI's announcement, underscoring how closely the two companies are tracking each other.

My read is that Sol and Luna are less a new capability statement than a pricing and efficiency response. Astra established the top of the GPT-6 range; these models are designed to make the generation easier to deploy at scale. The competitive claims will attract attention, but the API price cut is the part developers can test directly.

The models are already available in ChatGPT Work and Codex for most paid accounts, as well as through the ChatGPT API. Luna is also available in the desktop app and to Free and Go users. OpenAI plans to roll both models out across the ChatGPT app and website during the day.

That distribution makes the release a practical expansion of GPT-6, but its real test is whether lower inference costs turn Sol and Luna from upgraded models into default choices for everyday workloads.