OpenAI launches GPT-6 Sol and Luna with lower costs and fewer mistakes

OpenAI has launched GPT-6 Sol and Luna, adding updated smaller models to the GPT-6 generation after introducing GPT-6 Astra earlier this month. The company described Astra as its most powerful and capable model yet, and says Sol and Luna extend that generation’s benefits in a more efficient and accessible form.

The launch positions Sol and Luna as part of a broader model hierarchy rather than direct replacements for every GPT-6 use case. For users and developers, the practical question is which model tier fits the job: complex technical work, high-volume clerical tasks, or general ChatGPT use.

OpenAI says Sol is designed for complex tasks such as coding. Luna is positioned for clerical work, including “high-volume tasks with a clear goal,” such as summarizing documents, extracting information, or answering quick questions.

That distinction matters for teams choosing between capability and throughput. The outline from OpenAI suggests Sol is the model to evaluate for demanding workflows, while Luna is aimed at repetitive, structured tasks where speed and volume may matter more.

OpenAI is emphasizing efficiency and affordability for the new Sol and Luna releases. The company says the GPT-6 series versions will be available at half the cost of the 5.6 series of Sol and Luna for API access, attributing the price drop to improvements in caching and inference.

OpenAI also claims the models are better at producing factually accurate answers and have a lower error rate for coding. In its announcement, the company says GPT-6 Sol makes about half as many mistakes as its predecessor on an internal factuality evaluation based on de-identified real-world conversations where users flagged model errors.

The new Sol and Luna models are 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.

OpenAI expects to roll out the models to ChatGPT’s app and website gradually throughout the day. The launch also lands in a competitive moment: TechCrunch notes Anthropic released a new version of Opus 5.5 just 90 minutes before OpenAI’s release.