OpenAI lowers GPT-6 Sol and Luna prices

The updated models are rolling out across ChatGPT, Codex and the API, with OpenAI promising lower costs and fewer factual and coding errors.

OpenAI has expanded its GPT-6 lineup with updated versions of its smaller Sol and Luna models, putting fresh emphasis on cost, accuracy and broader availability. The company says the release builds on GPT-6 Astra, which it introduced earlier this month as its most powerful model so far. According to IT-PUB News, the update pairs lower API pricing with claims of fewer factual and coding mistakes. That could make it relevant both for businesses automating more work and for people using AI tools for writing, coding or everyday tasks.

OpenAI says the new models are cheaper to run

OpenAI says the new Sol and Luna models are meant to extend the benefits of GPT-6 Astra by making them “more efficient and accessible.” The company attributes the lower price to improvements in caching and inference, two parts of the process that affect how AI models handle requests and generate answers.

The clearest change is the price cut. OpenAI says the 6 series models will cost half as much as the 5.6 series of Sol and Luna when used through the API.

For companies and developers using OpenAI models at scale, that is a meaningful shift. Even modest pricing changes can affect how often a service is used and which tasks are handed over to it.

Sol and Luna target different kinds of tasks

OpenAI introduced Sol and Luna earlier this year as different tiers in its model lineup. The company presents Sol as the better fit for more complex work, including coding.

Luna, by contrast, is positioned for clerical and high-volume tasks. OpenAI says it is designed for jobs with a clear goal, such as summarizing documents, extracting information or answering quick questions.

That split shows how OpenAI is trying to steer different users toward different tools and workflows. Rather than treating every model as a general-purpose option, the company is separating them by task and expected workload.

For businesses, that may make it easier to choose a model based on cost and use case. For regular users, it could mean that some responses and features are better suited to routine administrative work, while others are aimed at more demanding tasks.

OpenAI says Sol makes fewer mistakes

OpenAI is also stressing what it describes as better factual accuracy. The company says the new models are more likely to produce correct answers and have a lower error rate in coding.

In one internal evaluation, based on de-identified real-world conversations where users flagged mistakes, OpenAI says GPT-6 Sol made about half as many mistakes as its predecessor. The company says that brought Sol to Astra-level reliability at a much lower cost.

Those claims matter because errors remain one of the biggest limits of AI tools. A model can sound confident and still return wrong or incomplete answers. By focusing on fewer mistakes, OpenAI is addressing one of the most persistent concerns around using AI in work settings.

Still, the results come from OpenAI’s own internal testing, not an independent assessment.

OpenAI again draws a contrast with Anthropic

As with many of its model launches, OpenAI used the announcement to compare its products with Anthropic’s. The company says GPT-6 Sol and Luna outperform Anthropic’s top models, including Fable and Opus, across a range of tasks.

The timing added another element to the release. Anthropic launched a new version of Opus 5.5 just 90 minutes before OpenAI made its announcement, underscoring how closely the two companies are competing.

That rivalry matters beyond the companies themselves. The pace of releases can shape which tools developers pick, how quickly businesses switch providers and how aggressively prices move across the AI market.

The updated models are rolling out across OpenAI products

OpenAI says the new versions of Sol and Luna are already available in ChatGPT Work and Codex for most paid accounts, as well as in the ChatGPT API. Luna will also be available in the desktop app and for Free and Go users.

The company says it will roll out the models gradually to ChatGPT on the app and website throughout the day.

That suggests a phased launch rather than a single switch for everyone at once. For users, access may depend on account type and timing. For OpenAI, it is another step in spreading the updated models across both consumer and professional products.

The message behind the release is fairly direct. OpenAI wants its smaller GPT-6 models to be seen not just as capable, but also cheaper and more dependable — and that matters most if the performance holds up outside the company’s own tests.