OpenAI has released GPT-6 Sol and GPT-6 Luna, expanding its GPT-6 model family with two lower-priced options aimed at coding, agentic workflows and high-volume work. The company says both models use methods similar to those behind GPT-6 Astra, its higher-end model.
According to OpenAI’s launch announcement, GPT-6 Sol is priced at $2 per million input tokens and $10 per million output tokens through the API. GPT-6 Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens. OpenAI compares those rates with GPT-5.6’s promotional pricing of $4 and $20 for Sol, and $0.20 and $1.20 for Luna, describing both new models as 50% cheaper on that basis.
Two models for different workloads
OpenAI positions Sol as the middle option in the family for complex coding and agentic work, while Luna is meant for focused, high-volume jobs where cost matters more. The company says the two models bring improvements in professional work, factuality, coding, computer use and alignment relative to their GPT-5.6 counterparts.
Those performance comparisons should be read as company evaluations, not independent benchmarks. OpenAI’s announcement cites results from its own factuality and safety testing alongside third-party benchmark names, but the tests, effort settings and cost assumptions vary by comparison. The practical differences will depend on the tools, prompts and workloads a developer actually runs.
The company also says it has improved prompt caching for GPT-6, with discounts of 90% on cached input-token reads. That matters most for agents and long-running applications that repeatedly send shared context: a cheaper model rate is only part of the bill when a system is processing a large reused prompt.
Where the models are available
Sol and Luna are available as gpt-6-sol and gpt-6-luna in the API, OpenAI says. The company says they are also available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Free and Go users can access Luna in the desktop app.
OpenAI says the models are not yet available in Chat and that its ChatGPT rollout will happen gradually. That distinction is important for people who see the launch announcement but do not immediately find either model in every ChatGPT surface or workspace.
