OpenAI Releases GPT-6 Sol and Luna With API Prices Half Their Predecessors’
The release creates two cheaper options below GPT-6 Astra, but OpenAI’s promised accuracy gains remain based on its own tests.
Listen to this story
The audio brief
Story brief
3 key pointsOpenAI is expanding its GPT-6 lineup downward from Astra with Sol and Luna, emphasizing lower operating costs rather than maximum capability. API rates are $2/$10 per million input/output tokens for Sol and $0.10/$0.50 for Luna, with OpenAI claiming 50% savings versus corresponding GPT-5.6 models. Luna reportedly scored 5.4% higher on AutomationBench at 58% lower cost per task, though those are vendor benchmarks....
- 01
Sol costs $2 input and $10 output per million tokens; Luna costs $0.10 and $0.50.
- 02
OpenAI reports both models make half as many factual errors as their predecessors.
- 03
Luna reportedly beat GPT-5.6 by 5.4% on AutomationBench while costing 58% less per task.
OpenAI has released GPT-6 Sol and GPT-6 Luna, two lower-cost models positioned as less-intensive alternatives to GPT-6 Astra. The company says both bring Astra-style training methods to faster, more affordable models for work at different scales.
Sol and Luna became available September 22 in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu customers. Luna is also available to Go and Free users in the ChatGPT desktop app. Neither model had reached basic Chat at publication.
The new price split
The clearest cost difference appears in the rate card. OpenAI lists Sol at $2 per million input tokens and $10 per million output tokens. Luna costs $0.10 for input and $0.50 for output at the same volume. OpenAI says both models cost 50% less than their corresponding GPT-5.6 versions.
OpenAI’s case rests on lower-cost accuracy
OpenAI’s pitch is not simply that Sol and Luna are cheaper. The company says the models make half as many factual errors as their predecessors. It also reports that Luna outperformed the GPT-5.6 release by 5.4% on AutomationBench while costing 58% less per task.
Those figures are OpenAI-reported benchmark and cost results, rather than a guarantee for a particular deployment. A lower rate can reduce spending, but a team’s outcome will still depend on the work it assigns and the model’s results in that setting.
What OpenAI says has changed
- Sol and Luna are positioned collectively as less-intensive alternatives to GPT-6 Astra.
- OpenAI says they use training methods similar to Astra’s.
- The company positions the pair for work at different scales.
A familiar pair in a new lineup
OpenAI previously described GPT-5.6 Sol as the flagship of its earlier series, with Luna as a fast, affordable model. The GPT-6 release places both below Astra instead, making them part of a broader set of choices organized around capability and cost.
Access is still uneven
The initial rollout favors paid Work and Codex customers. Free and Go users receive Luna only through the desktop app, while basic Chat had neither model at publication. That leaves access to the new pair dependent on product route and subscription tier.
Editorial analysis
Our Read
The meaningful change is not just a cheaper pair of models. OpenAI is making price differences within its lineup more explicit, while tying its quality case to company-reported factuality and benchmark results. That puts more pressure on customers to decide which tasks deserve a more expensive model rather than treating a model family as one product. The next useful signal will be how these models perform in actual workflows, especially where output volume and corrections affect the total cost of a task. Their absence from basic ChatGPT also keeps the launch focused, for now, on selected paid and desktop routes.
Sources
- openai.comPreviewing GPT-5.6 Sol: a next-generation model
- mashable.comChatGPT-6 Sol and Luna are here: Pricing and benchmarks data
Reader comments
Newest comments first. Replies stay oldest first.