SAN FRANCISCO: OpenAI launched GPT-6 Sol and Luna on September 22, expanding its GPT-6 model family with faster, lower-cost options built on advances introduced with GPT-6 Astra.
OpenAI said improved inference and caching allowed it to reduce prices compared with promotional GPT-5.6 rates. GPT-6 Sol costs USD 2 per million input tokens and USD 10 per million output tokens under standard API pricing.
GPT-6 Luna costs USD 0.10 per million input tokens and USD 0.50 per million output tokens. OpenAI describes Luna as its most efficient GPT-6 model for focused, high-volume tasks.
OpenAI positioned GPT-6 Sol between its flagship Astra model and the lower-cost Luna tier. In addition, the company said both models inherit advances in professional work, coding, factuality, computer use and alignment from Astra.
In OpenAI’s AutomationBench evaluation, GPT-6 Sol at xhigh reasoning effort scored 33.2% at USD 0.27 per task. OpenAI reported a 26.9% score for Claude Opus 5 at maximum effort in the same evaluation.
These figures come from OpenAI’s testing and should be read as vendor-reported benchmark results. Meanwhile, the company reported a 68.8% DeepSWE score for GPT-6 Sol at maximum effort.
GPT-6 Luna scored 66.6%. OpenAI said those tests measure performance on long-horizon software-engineering tasks in real codebases.
OpenAI also upgraded prompt caching for GPT-6. Cached input reads receive a 90% discount. New controls also let developers adjust reasoning effort and tools without automatically breaking earlier cached context.
GPT-6 Astra, Sol, and Luna: For production agents in Microsoft Foundry. Explore GPT-6 Astra, Sol, and GPT-6 Luna in Microsoft Foundry, with scalable model options for production AI agents, complex workflows, and high-volume tasks.
GPT-6 Sol and Luna are available through the OpenAI API as gpt-6-sol and gpt-6-luna. OpenAI also began rolling them out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Free and Go users can access GPT-6 Luna through the desktop app.
