OpenAI announced the launch of new models, Sol and Luna, in its GPT-6 series, aiming to improve efficiency and accessibility. These models will be available via API at half the price of the GPT-5.6 series Sol and Luna, which the company attributes to improvements in caching and inference. OpenAI also claims that in internal factual accuracy evaluations, GPT-6 Sol's error rate was reduced by approximately 50% compared to its predecessor, reaching the reliability level of the Astra model while being more cost-effective. Additionally, the new models show a reduced error rate in coding. The new versions of Sol and Luna are now available to most paid accounts on ChatGPT Work and Codex, and can be accessed via the ChatGPT API; Luna will also support desktop applications, as well as free and Go users. OpenAI expects to gradually roll out these models to the ChatGPT app and website later today. Notably, Anthropic also released its new Opus 5.5 ninety minutes before OpenAI's announcement, highlighting the fierce competition between the two companies.