OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes

OpenAI has unveiled the latest iterations of its GPT-6 generation, dubbed Sol and Luna, promising enhanced efficiency and affordability. According to the company, GPT-6 Astra, previously announced, marked the advent of a new level of intelligence. These new models, Sol and Luna, aim to make that intelligence more efficient and accessible by extending the benefits of Astra.

Sol is tailored for intricate tasks such as coding, while Luna is more suited for clerical work involving high-volume tasks with clear objectives, including summarizing documents, data extraction, and quick answers to queries. OpenAI has also highlighted significant reductions in pricing, with the GPT-6 series available at half the cost of the preceding 5.6 series. This price cut is attributed to advancements in caching and inference processes.

The company has also asserted improvements in answer accuracy and a lower error rate in coding tasks. Specifically, GPT-6 Sol has reportedly made half as many mistakes as its predecessor, achieving Astra-level reliability at a much lower cost. OpenAI continues to assert its models surpass those of its main competitor, Anthropic, particularly in tasks like those handled by Anthropic's top models Fable and Opus.

The new Sol and Luna models are currently accessible via ChatGPT Work and Codex to most paid subscribers, and through the ChatGPT API. While Luna will be available in the desktop app and for Free and Go users. OpenAI anticipates a phased rollout of the models across ChatGPT, both on the app and website, across the day.