OpenAI has introduced updated versions of its GPT-6 Sol and GPT-6 Luna models, extending the company's latest AI generation with a focus on lower operating costs, stronger reliability and broader accessibility.
The two models serve different professional needs. Sol is positioned for demanding workflows such as software development and complex reasoning, while Luna is built for high-volume, clearly defined tasks including document summaries, information extraction and rapid question answering.
Lower Costs for AI Development
According to OpenAI, the new GPT-6 series will be offered through its API at around half the cost of the previous Sol and Luna generation. The company attributes this improvement to advances in caching and inference, which could help teams deploy capable AI tools across more products and services.
OpenAI also reports gains in factual reliability and coding performance. In its internal evaluations based on de-identified user conversations where model errors were flagged, GPT-6 Sol reportedly made roughly half as many factual mistakes as its predecessor while delivering reliability closer to the higher-end Astra model.
Wider Access Across OpenAI Tools
The updated models are becoming available through ChatGPT Work, Codex and the ChatGPT API for most paid users. Luna is also planned for the desktop application, as well as Free and Go users, with a broader rollout across ChatGPT continuing in stages.
By combining lower prices with improvements in accuracy, GPT-6 Sol and Luna may support a future where advanced AI becomes a more practical everyday tool for creators, developers and knowledge-driven teams.
