OpenAI introduced GPT-6 Sol and GPT-6 Luna on September 22, 2026, expanding its GPT-6 model family with two options built around different needs. Sol targets demanding work that benefits from stronger reasoning, while Luna focuses on efficient, high-volume tasks.
The launch follows the introduction of GPT-6 Astra, the flagship model in the same family. Together, the three models give developers and eligible ChatGPT users more flexibility when choosing how much capability and computing cost a task requires.
For people who use AI to write, research, build software or automate workflows, the important question is not simply which model is newest. It is which model fits the work they want to complete.
What Are GPT-6 Sol and GPT-6 Luna?
GPT-6 Sol and GPT-6 Luna are reasoning models in OpenAI’s GPT-6 family. Both accept text and image inputs and generate text responses through supported API endpoints.
Their main difference is the balance between capability and efficiency.
GPT-6 Sol is designed for complex coding and agentic workflows. It is intended for tasks that may involve several steps, tools, decisions and revisions.
GPT-6 Luna is designed for focused, high-volume workloads where efficiency and operating cost are especially important.
OpenAI positions GPT-6 Astra as its flagship model for the most demanding work, with Sol and Luna offering different options for projects that do not always require that level of capability.
Read OpenAI’s GPT-6 Sol and Luna announcement
GPT-6 Sol: Designed for Demanding Work
GPT-6 Sol is aimed at developers and professionals working on tasks that require more than a quick, straightforward response.
Examples of work to explore with Sol include:
- Building or updating software across multiple files.
- Planning and carrying out multi-step workflows.
- Working with tools to retrieve information or complete supported actions.
- Reviewing complex technical material and producing structured results.
These are examples of the kinds of tasks Sol is designed to support, not a guarantee that every result will be correct. Code, calculations and consequential decisions still require appropriate human review.
For developers, Sol is available through the OpenAI API under the model ID gpt-6-sol.
Explore the GPT-6 Sol model documentation
GPT-6 Luna: Built for Efficient, High-Volume Tasks
GPT-6 Luna takes a different approach. OpenAI describes it as an efficient model for focused tasks that may need to be performed repeatedly or at scale.
Potential uses include drafting and categorising content, extracting information into a consistent format, assisting with routine application features and handling large volumes of similar requests.
Its lower API price may be especially relevant to developers whose applications process many requests. However, a lower price per token does not automatically make every task cheaper: the total cost also depends on how many tokens, tool calls and attempts a workflow needs.
Luna is available in the API under the model ID gpt-6-luna.
Explore the GPT-6 Luna model documentation
GPT-6 Sol vs. Luna: What Is the Difference?
The two models share a model family, but they serve different workload profiles.
Choose what to test based on the task. If your workflow involves complex coding, extended reasoning or several coordinated steps, Sol is the model OpenAI positions for that type of work. If you need to process a large volume of focused requests at a lower per-token cost, Luna is designed for that use case.
The practical way to compare them is to run the same representative tasks through both models. Review response quality, completion time, token use and total cost—including any tool charges—before choosing one for a production application.
There is no need to use the same model for every part of a project. A developer might use one model for complex planning and another for simpler repeated operations.
Where Are GPT-6 Sol and Luna Available?
Availability depends on which OpenAI product you use, as well as your subscription or workspace permissions.
According to OpenAI’s September 22 announcement, GPT-6 Sol and Luna began rolling out in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users.
OpenAI also announced that Free and Go users can try GPT-6 Luna in the desktop app.
Both models are available through the OpenAI API for developers with the appropriate access.
An important distinction: availability in ChatGPT Work or Codex is not the same as availability in a standard ChatGPT conversation. OpenAI’s release notes describe the Work and Codex models as separate from the models available in Chat. If a model does not appear where you expect it, check the relevant product, your plan, workspace settings and current rollout status.
Check the latest ChatGPT release notes
How Much Do GPT-6 Sol and Luna Cost in the API?
For developers, the API pricing illustrates the difference between the models.
OpenAI’s published standard text-token rates for prompts of up to 272,000 input tokens are listed below. All prices are in US dollars and are calculated per one million tokens.
GPT-6 Sol API Pricing
- Input tokens: $2.00 per 1 million tokens.
- Cached input tokens: $0.20 per 1 million tokens.
- Output tokens: $10.00 per 1 million tokens.
GPT-6 Luna API Pricing
- Input tokens: $0.10 per 1 million tokens.
- Cached input tokens: $0.01 per 1 million tokens.
- Output tokens: $0.50 per 1 million tokens.
At these rates, Luna has a substantially lower per-token price than Sol. OpenAI also says that the launch reduced Sol and Luna API prices by 50% compared with GPT-5.6 promotional pricing.
These figures are API prices, not ChatGPT subscription prices. They also do not represent a guaranteed total cost for an application. Longer prompts, cache writes, tool usage and other processing options can have separate pricing rules.
Before deploying either model, check the current pricing page and estimate costs using your application’s actual request patterns.
Review the official OpenAI API pricing
What Can Developers Build with the New Models?
Both Sol and Luna support text and image input, text output and developer workflows through the Responses API.
OpenAI’s model documentation also lists support for capabilities such as function calling, structured outputs and supported tools. This gives developers ways to build applications that do more than return a paragraph of text.
For example, a business could explore an AI assistant that reads submitted information, organises it into a structured format and helps prepare a response for an employee to review. A software team could test an agent that inspects a project, proposes changes and runs supported development tools.
The model is only one part of such a system. Developers still need to design permissions, handle errors, test reliability and decide which actions require human approval.
Does This Mean Everyone Should Switch to GPT-6?
Not necessarily. A new model launch is a useful opportunity to revisit your tools, but the right choice depends on the tasks you perform and the products you can access.
If you primarily use ChatGPT for everyday questions, the models available in your normal chat may already suit your needs. If you use Work or Codex for larger projects, the new options may be worth testing when they become available to your account.
For API developers, a controlled comparison is particularly valuable. Use a small set of real tasks, establish what a successful result looks like and measure quality alongside speed and cost.
That approach will tell you more about the value of Sol or Luna for your own workflow than a model name alone.
Frequently Asked Questions
When were GPT-6 Sol and GPT-6 Luna released?
OpenAI announced the two models on September 22, 2026.
Are GPT-6 Sol and Luna the same model?
No. They belong to the same GPT-6 family but are designed for different balances of capability and cost. Sol targets more demanding workflows, while Luna emphasises efficiency for focused, high-volume tasks.
Can I use GPT-6 Sol and Luna in a normal ChatGPT conversation?
Do not assume that access in Work or Codex means both models are available in standard ChatGPT chat. Check the model options shown in your account and OpenAI’s current release notes.
Are the models available through the OpenAI API?
Yes. Developers can use the model IDs gpt-6-sol and gpt-6-luna, subject to applicable API access and usage limits.
Do GPT-6 Sol and Luna generate videos?
Their documented API model capabilities include text and image inputs with text output. They should not be confused with dedicated video-generation models or tools.
Final Thoughts
The arrival of GPT-6 Sol and GPT-6 Luna expands the choices available within OpenAI’s latest model family.
Sol is positioned for demanding coding and agentic tasks. Luna offers an efficiency-focused option for applications that need to handle many focused requests. Both give developers additional ways to match a model to the work being done.
For ChatGPT users, the most important detail is where the models are available: the September launch specifically identifies Work, Codex and the API, with access varying by plan and product.
For developers, the next step is straightforward: test the models on real workloads, compare results and review the full cost before deciding how to use them.
