Earlier this month, OpenAI launched GPT-6 Astra, heralded as the world’s best model for a variety of tasks, including computer work and coding. Now, the lab has expanded the GPT-6 line with updated versions of the Sol and Luna models, boasting lower costs and fewer mistakes.

The first models in the Sol and Luna series, designed for complex coding tasks and high-volume clerical work respectively, have been refined. GPT-6 Sol, in particular, has halved the mistake rate compared to its predecessor, reaching Astra-level reliability at a much lower cost. OpenAI attributes this to significant improvements in caching and inference.

Efficiency and affordability are the key selling points, with the 6 series models available at half the cost of the 5.6 series. OpenAI also claims that its newest models are better at producing factually accurate answers, with a lower error rate when it comes to coding. However, the company is also quick to claim that its offerings outshine those of its primary competitor, Anthropic, with GPT-6 Sol and Luna handling tasks substantially better than Fable and Opus.

The new versions of Sol and Luna are now available in ChatGPT Work, Codex, and the ChatGPT API, with Luna also available in the desktop app and for Free and Go users. OpenAI expects to roll out the models to ChatGPT gradually throughout the day, reflecting the ongoing competition with Anthropic, which released a new version of Opus 5.5 just 90 minutes prior to OpenAI’s release.