OpenAI has expanded its GPT-6 model family with the launch of GPT-6 Sol and GPT-6 Luna, promising lower costs, improved efficiency and fewer errors compared with their predecessors.
The release follows the launch of GPT-6 Astra earlier this month, which OpenAI described as its most powerful model yet for tasks including coding and computer-based work.
According to OpenAI, the new Sol and Luna models bring some of Astra’s capabilities to more efficient and accessible systems.
Sol targets complex tasks
The Sol and Luna models represent different capability tiers within OpenAI’s model lineup.
GPT-6 Sol is designed for more complex tasks, including coding and advanced knowledge work, while Luna is aimed at high-volume tasks with clearly defined objectives.
OpenAI says Luna is particularly suited to activities such as summarising documents, extracting information and answering routine questions.
OpenAI cuts API prices
One of the main changes in the new models is their cost.
OpenAI says the GPT-6 versions of Sol and Luna will be available through its API at half the cost of their GPT-5.6 predecessors.
The company attributed the price reduction to improvements in caching and inference efficiency.
The lower costs could make the models more attractive for developers and businesses running AI systems at scale.
New models claim fewer mistakes
OpenAI also highlighted improvements in factual accuracy and coding reliability.
The company said GPT-6 Sol makes roughly half as many mistakes as its predecessor on an internal factuality evaluation based on de-identified real-world conversations in which users had flagged model errors.
OpenAI said the results bring Sol to a level of reliability comparable with GPT-6 Astra, while operating at a substantially lower cost.
The company also reported improvements in coding accuracy, although the performance figures are based on OpenAI’s own evaluations.
OpenAI and Anthropic intensify AI race
The release comes amid intensifying competition between OpenAI and Anthropic.
OpenAI claims GPT-6 Sol and Luna outperform Anthropic’s leading models, including Fable and Opus, on a range of tasks.
Such comparisons are based on benchmarks and evaluations cited by the companies and may vary depending on the task, testing methodology and model configuration.
The timing of the release also highlights the increasingly rapid pace of competition. Anthropic released its latest Opus 5.5 model around 90 minutes before OpenAI announced GPT-6 Sol and Luna.
GPT-6 Sol and Luna availability
The new models are available through ChatGPT Work and Codex for most paid accounts, as well as through the ChatGPT API.
Luna is also being made available in the desktop app and to Free and Go users.
OpenAI said it would gradually roll out GPT-6 Sol and Luna across the ChatGPT app and website.
