What Is GPT-6 Luna? Pricing, Features, Context Window & How to Use It
What Is GPT-6 Luna?
GPT-6 Luna is OpenAI’s newest efficiency-focused model in the GPT-6 family, designed for focused tasks that need to run at high volume without the higher per-token cost of larger frontier models.
OpenAI released GPT-6 Luna on September 22, 2026, alongside GPT-6 Sol. The company describes Luna as its most efficient GPT-6 model for focused, high-volume workloads.
GPT-6 Luna combines an unusually large 1,050,000-token context window with up to 128,000 output tokens, image input, reasoning capabilities, structured outputs, function calling and support for tools such as web search, file search and computer use.
Its most striking feature, however, may be its price.
OpenAI currently lists standard short-context pricing at $0.10 per million input tokens and $0.50 per million output tokens, making GPT-6 Luna significantly cheaper than GPT-6 Sol and GPT-6 Astra.
That combination makes Luna particularly relevant for developers and businesses building repeatable AI workflows, classification systems, document-processing pipelines, support automation and AI agents for small business.
This guide explains exactly what GPT-6 Luna is, how much it costs, how its context window works, what tools it supports, how it compares with GPT-6 Sol and when it makes sense to use it.
What Is GPT-6 Luna?
GPT-6 Luna is part of OpenAI’s GPT-6 model family.
The current family consists of three primary models:
- GPT-6 Astra for the hardest end-to-end reasoning, coding, research and computer-use work.
- GPT-6 Sol for complex coding and agentic workflows with a stronger balance between capability and cost.
- GPT-6 Luna for efficient, focused and high-volume tasks.
This positioning is important.
Luna is not intended to replace every larger model.
Instead, it is designed for situations where an application may need to process very large numbers of requests efficiently.
That makes it particularly interesting for businesses operating AI at scale.
Examples could include document classification, customer-query routing, structured extraction, large-scale summarization, automated research steps and repetitive agent workflows.
GPT-6 Luna Release Date
OpenAI released GPT-6 Luna on September 22, 2026.
The official API changelog confirms the launch of:
gpt-6-luna
alongside:
gpt-6-sol
Both models support text and image input and can generate text through OpenAI’s Responses API and Chat Completions API.
OpenAI also introduced GPT-6 Sol and Luna into ChatGPT Work and Codex on September 22. The ChatGPT release notes specify that these models are available there separately from the models used in regular Chat.
GPT-6 Luna Specifications
Here are the key GPT-6 Luna specifications currently published by OpenAI.
| Feature | GPT-6 Luna |
|---|---|
| Release date | September 22, 2026 |
| Model ID | gpt-6-luna |
| Context window | 1,050,000 tokens |
| Maximum output | 128,000 tokens |
| Standard input price | $0.10 / 1M tokens |
| Cached input price | $0.01 / 1M tokens |
| Standard output price | $0.50 / 1M tokens |
| Knowledge cutoff | May 18, 2026 |
| Text input | Supported |
| Image input | Supported |
| Text output | Supported |
| Reasoning | Supported |
| Structured outputs | Supported |
| Function calling | Supported |
OpenAI’s current model documentation provides these specifications and identifies Luna as the efficiency-oriented member of the GPT-6 family.
GPT-6 Luna Pricing
GPT-6 Luna is one of the most aggressively priced models in the current GPT-6 family.
For standard processing with shorter contexts, OpenAI currently lists:
Input: $0.10 per 1 million tokens
Cached input: $0.01 per 1 million tokens
Cache writes: $0.125 per 1 million tokens
Output: $0.50 per 1 million tokens
This pricing can be especially important for applications processing thousands or millions of routine requests.
However, developers should understand that context length affects pricing.
Does GPT-6 Luna Cost More for Long Contexts?
Yes.
OpenAI’s pricing documentation distinguishes between short-context and long-context requests.
For standard GPT-6 Luna processing, OpenAI currently lists:
| Context Pricing | Input | Cached Input | Cache Writes | Output |
|---|---|---|---|---|
| Short context | $0.10 | $0.01 | $0.125 | $0.50 |
| Long context | $0.20 | $0.02 | $0.25 | $0.75 |
Prices are per one million tokens.
This means developers using Luna primarily because of its enormous context window should model costs using the appropriate long-context pricing rather than assuming the lowest advertised rate applies to every request.
GPT-6 Luna Context Window
GPT-6 Luna supports a 1,050,000-token context window.
That is more than one million tokens of working context.
A large context window allows developers to provide substantial amounts of information to the model during the same request or workflow.
Potential applications include:
- Large collections of documents
- Long business reports
- Extensive knowledge bases
- Large code repositories
- Customer histories
- Research collections
- Long conversations
- Technical documentation
- Multiple related files
A large context window does not automatically guarantee perfect understanding of every token.
However, it gives applications considerably more room to supply relevant information without breaking material into many disconnected requests.
GPT-6 Luna Maximum Output
GPT-6 Luna supports up to 128,000 output tokens.
This is useful for workloads requiring substantial generated output, including long structured documents, code generation, large transformations and comprehensive reports.
The maximum output limit should not be confused with the context window.
The context window represents how much information can fit into the model’s working context.
The maximum output describes how much the model can generate in response.
GPT-6 Luna Reasoning Levels
GPT-6 Luna supports reasoning tokens and multiple reasoning-effort settings.
OpenAI currently documents the following options:
none
low
medium
high
xhigh
max
The default reasoning setting is medium.
This allows developers to adjust how much computational effort the model applies to a task.
For example, a simple classification workflow may not require the same reasoning effort as a complicated multi-step analysis.
That flexibility can help developers balance three factors:
Capability, latency and cost.
What Tools Does GPT-6 Luna Support?
GPT-6 Luna is not limited to plain text generation.
OpenAI’s current model catalog lists support for tools including:
- Function calling
- Web search
- File search
- Computer use
The broader GPT-6 model guidance also describes features such as asynchronous tool calling, which allows GPT-6 models to continue reasoning or work on other parts of a task while an application executes a tool call.
This is particularly relevant for agentic applications.
Instead of simply generating an answer, an AI system can potentially decide which tool it needs, call that tool and continue working with the returned information.
That is the same broader shift discussed in our guide to AI agents for small business.
GPT-6 Luna and AI Agents
GPT-6 Luna could be especially interesting for AI-agent systems where a large number of relatively focused steps need to be executed economically.
Consider a customer-service workflow.
An AI agent might need to:
- Understand a customer request.
- Categorize the request.
- Search a knowledge base.
- Retrieve account information.
- Generate a response.
- Decide whether human escalation is necessary.
Using the largest and most expensive frontier model for every individual step may not always be economical.
An efficiency-focused model such as Luna can potentially handle many routine parts of an agent workflow while larger models are reserved for more difficult decisions.
Businesses exploring this type of automation may also want to read our guide to AI customer service agents for small business.
GPT-6 Luna for Customer Service
Customer support is a strong example of where high-volume AI economics matter.
A business may receive hundreds or thousands of customer requests containing repetitive questions.
Possible Luna-powered tasks could include:
- Request classification
- Knowledge-base retrieval
- Conversation summarization
- Intent detection
- Ticket routing
- Structured data extraction
- Draft responses
- Escalation detection
For customer-facing deployment, businesses should still monitor accuracy and provide appropriate human escalation.
A low token price does not remove the risks associated with incorrect customer information.
For a deeper explanation of these workflows, see AI customer service agents for small business.
GPT-6 Luna for AI Receptionists
Voice-based AI systems often involve several components.
One system converts speech into text.
Another model understands the request and determines what should happen.
Additional tools may check availability, retrieve customer information or schedule appointments.
A model optimized for repeated high-volume tasks could therefore become part of an AI receptionist workflow.
Businesses interested specifically in phone automation can read our guide to AI receptionist for small business.
GPT-6 Luna for Document Processing
The combination of low token pricing and a large context window makes document-processing workflows an obvious use case to evaluate.
Examples include:
- Invoice extraction
- Contract categorization
- Document classification
- Report summarization
- Customer-record analysis
- Knowledge-base ingestion
- Form processing
- Data normalization
The most important consideration is not simply whether Luna can process the document.
Developers should test whether its accuracy is sufficient for the specific workflow.
For high-stakes documents involving legal, financial, medical or other consequential information, independent verification remains important.
GPT-6 Luna for Coding
GPT-6 Luna can work with coding-related tasks, but OpenAI positions GPT-6 Sol more specifically around complex coding and agentic workflows.
That distinction can help developers choose between them.
GPT-6 Luna may make sense for high-volume coding-related tasks such as:
- Basic transformations
- Code classification
- Documentation
- Repetitive code generation
- Test generation
- Simple debugging
- Code summarization
More difficult repository-wide engineering work may justify a stronger model such as GPT-6 Sol or GPT-6 Astra.
If advanced coding capability is your priority, another current frontier model worth examining is Claude Opus 5.5, which Anthropic specifically positions for long-running agentic coding.
GPT-6 Luna vs GPT-6 Sol
GPT-6 Luna and GPT-6 Sol share several important technical characteristics.
Both have:
- 1,050,000-token context windows
- 128,000-token maximum outputs
- Text input
- Image input
- Reasoning support
- Function calling
- Tool support
The main difference is positioning and price.
| Feature | GPT-6 Luna | GPT-6 Sol |
|---|---|---|
| Main positioning | Focused high-volume work | Complex coding and agents |
| Standard input | $0.10 / 1M | $2 / 1M |
| Standard output | $0.50 / 1M | $10 / 1M |
| Cached input | $0.01 / 1M | $0.20 / 1M |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
OpenAI describes Luna as its most efficient model for focused, high-volume tasks, while Sol is built for complex coding and agentic workflows.
That means the better model depends on the workload.
Use Luna when scale and efficiency are major priorities.
Consider Sol when a workflow needs stronger capability for complex coding or agentic reasoning.
GPT-6 Luna vs GPT-6 Astra
GPT-6 Astra sits at the highest end of OpenAI’s current GPT-6 lineup.
OpenAI describes Astra as its most capable model for the hardest end-to-end work, including complex reasoning, coding, computer use, research and document creation.
Standard pricing is also dramatically higher.
GPT-6 Astra currently costs:
$10 per million input tokens
and:
$50 per million output tokens
for standard short-context processing.
GPT-6 Luna starts at:
$0.10 per million input tokens
and:
$0.50 per million output tokens.
The difference reflects two different objectives.
Astra prioritizes maximum capability.
Luna prioritizes efficiency at scale.
GPT-6 Luna vs GPT-5.6 Luna
GPT-6 Luna also succeeds the efficiency-focused GPT-5.6 Luna model.
OpenAI lists GPT-5.6 Luna at $0.20 per million input tokens and $1.20 per million output tokens on its model page.
GPT-6 Luna is listed at:
$0.10 input
and:
$0.50 output
per million tokens for standard short-context processing.
OpenAI’s launch announcement says Sol and Luna benefit from more efficient caching and inference, with lower API prices than their GPT-5.6 counterparts.
Does GPT-6 Luna Support Images?
Yes.
OpenAI confirms that GPT-6 Luna accepts both text and image input while generating text output.
This allows developers to build workflows involving:
- Screenshots
- Documents
- Charts
- Photos
- Diagrams
- User interfaces
- Scanned information
The usefulness of vision depends on the individual application, so developers should test representative examples rather than assuming identical accuracy across all image types.
Does GPT-6 Luna Support Structured Outputs?
Yes.
OpenAI’s model comparison documentation lists structured outputs as supported by GPT-6 Luna.
This can be particularly valuable for automation.
Instead of receiving unstructured paragraphs, developers can require responses in predictable schemas that downstream software can process more reliably.
Potential applications include:
- Extracting invoice fields
- Producing JSON objects
- Categorizing support requests
- Populating databases
- Creating structured reports
- Generating API-ready output
GPT-6 Luna API
The model ID is:
gpt-6-luna
OpenAI supports GPT-6 Luna through both the Responses API and Chat Completions API.
For newer applications using built-in tools, OpenAI recommends the Responses API.
The GPT-6 Luna documentation states that built-in tools and function calling are available through the Responses API. Function calling through Chat Completions has additional restrictions depending on reasoning configuration.
How to Use GPT-6 Luna
For developers, the basic process is straightforward.
First, select:
gpt-6-luna
as the model in your OpenAI API request.
Then choose an appropriate reasoning-effort setting based on the complexity of the task.
For agentic workflows, use the Responses API when you need capabilities such as built-in tools.
Before deploying to production, test representative inputs and compare:
- Accuracy
- Latency
- Token consumption
- Cost
- Failure rate
- Escalation requirements
A cheaper model only creates value if it still performs the task reliably.
Is GPT-6 Luna Available in ChatGPT?
OpenAI’s September 22, 2026 ChatGPT release notes state that GPT-6 Luna and GPT-6 Sol were introduced into ChatGPT Work and Codex.
The release notes specifically note that these models are separate from the models available in regular Chat.
OpenAI’s developer announcement also states that the models rolled out to ChatGPT Work and Codex for eligible paid plans, while GPT-6 Luna could be tried by Free and Go users through the desktop app at launch.
Availability can change, so users should check their current model picker, Work environment or Codex interface rather than relying permanently on launch-day availability.
GPT-6 Luna Knowledge Cutoff
OpenAI currently lists GPT-6 Luna’s knowledge cutoff as:
May 18, 2026.
This means the model’s built-in knowledge should not automatically be assumed to include events occurring after that date.
Applications that need current information should provide an appropriate live information source such as web search, retrieval or another connected system.
Is GPT-6 Luna Good for Small Businesses?
It may be particularly attractive for small businesses when the AI workload is repetitive and high-volume.
Possible applications include:
- Customer-support triage
- Email classification
- Lead categorization
- Product-data processing
- Document extraction
- Internal search
- Report preparation
- Content tagging
- Workflow routing
- AI-agent sub-tasks
The important question is not simply whether GPT-6 Luna is inexpensive.
The business should measure whether it completes the job accurately enough.
For broader automation ideas, see our guide to AI agents for small business.
GPT-6 Luna and the Future of Agentic AI
The arrival of low-cost models with large context windows could change how AI agents are designed.
Instead of relying on one expensive model for every step, developers can build systems where different models handle different levels of work.
For example:
Luna → routine classification and retrieval
Sol → more difficult workflow decisions
Astra → the hardest reasoning problems
This model-routing approach can potentially reduce cost while preserving stronger models for situations where their additional capability matters.
Consumer products are moving in a similar direction.
Personal agents such as Meta Muse AI illustrate how AI is shifting from answering questions toward performing multi-step tasks on behalf of users.
You can also read our Meta Muse AI step-by-step guide for an example of how agentic AI is being presented to everyday users.
GPT-6 Luna Advantages
The strongest reasons to consider GPT-6 Luna are its low token price, large context window and access to modern GPT-6 capabilities.
Its combination of:
$0.10 standard short-context input pricing
$0.50 output pricing
1.05M context
128K maximum output
reasoning support
image input
structured outputs
and tool support
makes it potentially useful for large-scale AI workloads.
GPT-6 Luna Limitations
GPT-6 Luna is not positioned as OpenAI’s strongest model.
Complex reasoning, advanced coding or difficult end-to-end professional work may benefit from GPT-6 Sol or GPT-6 Astra.
Other limitations include the possibility of incorrect outputs, higher pricing for very long contexts and the need for external retrieval when current information falls beyond the May 18, 2026 knowledge cutoff.
Applications involving financial, legal, medical, security or other consequential decisions should not rely solely on an AI model without appropriate verification and safeguards.
Frequently Asked Questions About GPT-6 Luna
What is GPT-6 Luna?
GPT-6 Luna is OpenAI’s efficiency-focused GPT-6 model designed for focused, high-volume AI workloads. It supports reasoning, text and image input, structured outputs, function calling and a range of tools.
When was GPT-6 Luna released?
GPT-6 Luna was released on September 22, 2026.
How much does GPT-6 Luna cost?
For standard short-context API usage, GPT-6 Luna is listed at $0.10 per million input tokens and $0.50 per million output tokens.
What is the GPT-6 Luna context window?
GPT-6 Luna has a 1,050,000-token context window.
What is the GPT-6 Luna maximum output?
GPT-6 Luna supports up to 128,000 output tokens.
What is the GPT-6 Luna model ID?
The API model ID is:
gpt-6-luna
Does GPT-6 Luna support reasoning?
Yes. GPT-6 Luna supports reasoning effort settings ranging from none through max, with medium documented as the default.
Does GPT-6 Luna support images?
Yes. GPT-6 Luna accepts image input as well as text input.
Does GPT-6 Luna support web search?
OpenAI’s model catalog lists web search among the tools supported by GPT-6 Luna.
Does GPT-6 Luna support computer use?
Yes. Computer use is listed among its supported tools.
Is GPT-6 Luna cheaper than GPT-6 Sol?
Yes. OpenAI lists Luna at $0.10 per million standard short-context input tokens and $0.50 per million output tokens, while Sol is listed at $2 and $10 respectively.
Is GPT-6 Luna better than GPT-6 Sol?
They target different workloads. OpenAI positions Luna for efficient high-volume tasks and Sol for complex coding and agentic workflows. The better choice therefore depends on the task rather than model name alone.
Is GPT-6 Luna good for AI agents?
Its low cost, reasoning support, large context window and tool access can make it relevant for repeatable agentic workloads. More difficult agent tasks may benefit from stronger GPT-6 models.
Is GPT-6 Luna available in ChatGPT?
OpenAI introduced GPT-6 Luna in ChatGPT Work and Codex on September 22, 2026. The release notes state that these models are separate from regular Chat models.
What is GPT-6 Luna’s knowledge cutoff?
OpenAI lists the knowledge cutoff as May 18, 2026.
GPT-6 Luna: The Bottom Line
GPT-6 Luna is one of the most interesting GPT-6 releases for developers and businesses concerned with AI cost at scale.
It combines a 1.05-million-token context window, 128,000-token maximum output, reasoning support, image understanding, structured outputs and modern agent tools with pricing starting at $0.10 per million input tokens and $0.50 per million output tokens for standard short-context processing.
That does not make Luna the right choice for every workload.
OpenAI positions GPT-6 Astra for its hardest end-to-end tasks and GPT-6 Sol for demanding coding and agentic workflows. Luna instead targets the part of the AI market where efficiency, repetition and scale matter most.
For developers, that makes Luna worth evaluating anywhere an application makes large numbers of focused AI calls.
For businesses, it could make increasingly sophisticated automation economically practical.
To understand how these models fit into business automation more broadly, read our guide to AI agents for small business.
For customer-facing automation, continue with AI customer service agents for small business.
For phone-based automation, see AI receptionist for small business.
And if you want to compare the current agent trend with a consumer-focused product, read What Is Meta Muse AI?.
