OpenAI Efficient reasoning & automation
GPT-6 Luna
Handle frequent work with a smaller cost per useful result.
GPT-6 Luna is OpenAI's efficiency-oriented route for high-volume summaries, classification, rewriting and focused agent calls. TokenHot lists the same 1.05M context family, text and image input, reasoning, function calling and structured output, while positioning Luna as the lower-cost choice for work that does not need Sol's deepest long-horizon execution.
- High-volume reasoning
- 1.05M context
- Structured output
Luna is a route-selection option, not a promise that every task will match GPT-6 Sol. Keep a hard-task fallback and evaluate quality, correction rate and total cost on your own traffic.
AI Chat
Send a prompt and see the response here.
gpt-6-lunaStart a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|
| default | $0.1 | $0.5 | $0.01 | $0.125 |
| openai-0.2 | $0.02 | $0.1 | $0.002 | $0.025 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is GPT-6 Luna?
A good Luna workload has a clear answer shape and a cheap validation step: classify a ticket, summarize a report, rewrite a message or extract fields from an image.
Use the model family's large context deliberately, but avoid sending irrelevant material simply because the window is available. Smaller, well-scoped requests make cost and failure analysis easier.
Define the answer shape
Use a compact schema and a representative sample. Make invalid or uncertain items explicit rather than forcing a label.
Optimize for a quick useful pass
Give the model the few constraints that change the result. Reserve human review for low-confidence or high-impact cases.
Escalate deliberately
Use confidence, task type or validation failures to send hard cases to a deeper route such as GPT-6 Sol.
From brief to deliverable
Put GPT-6 Luna to work
Support and operations teams
Classify and route incoming requests
Provide category definitions, examples and a structured output schema. Ask the model to return an uncertainty reason when no category is a good fit.
- Tools you need
- A queue worker, schema validator and fallback reviewer.
- Keep in mind
- Do not hide ambiguity in a forced label; route unclear items for review.
View a task brief
Use the supplied a ticket, category schema, escalation rules and a destination queue. to plan classify and route incoming requests. Return a validated classification plus a routing reason., state the limits of the evidence and identify the checks required before accepting the result. Do not claim that an output or interaction was verified without direct evidence.
Analytics and operations teams
Summarize a recurring report set
Turn a batch of dated reports into short summaries with the same headings and a separate list of changes since the last period.
- Tools you need
- Document ingestion and a post-generation format check.
- Keep in mind
- A short summary can omit a critical exception; retain source links and spot-check high-impact changes.
View a task brief
Use the supplied reports, a stable summary schema, date range and audience. to plan summarize a recurring report set. Return consistent summaries with source pointers., state the limits of the evidence and identify the checks required before accepting the result. Do not claim that an output or interaction was verified without direct evidence.
Back-office and workflow teams
Extract fields from visual intake forms
Use image input for a first-pass extraction, then validate required fields and send low-confidence records to a human queue.
- Tools you need
- Image ingestion, schema validation and a human correction queue.
- Keep in mind
- Image understanding can miss small text, handwriting or cropped context; do not auto-approve critical fields.
View a task brief
Use the supplied a form image, field schema, allowed values and confidence policy. to plan extract fields from visual intake forms. Return structured fields plus a review flag., state the limits of the evidence and identify the checks required before accepting the result. Do not claim that an output or interaction was verified without direct evidence.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Context capacity | 1.05M tokens shown in the catalog | Use only the source needed for the decision so high-volume requests remain predictable. |
| Input and output | Text and image input; text output | Route visual extraction or analysis tasks without implying image generation. |
| Automation controls | Reasoning, function calling and structured output are listed | Use schemas, retries and escalation rules to turn a completion into a dependable workflow. |
| Cost position | Lower-cost family option than GPT-6 Sol | Use Luna as a default candidate for repeatable tasks, then escalate when validation shows the work needs more depth. |
| Version identity | TokenHot model ID: gpt-6-luna | Keep the exact model ID in batch logs and compare after catalog updates. |
Capabilities and price positioning come from TokenHot's current catalog. Use the route as an efficiency hypothesis and validate it against accepted output on your traffic.
Audience & fit
Where GPT-6 Luna fits
Teams processing repeated text work
Use Luna for summaries, rewrites, classification and extraction where the output schema and validation path are stable.
Operators building escalation queues
Make uncertainty and review states first-class so cost savings do not become silent quality loss.
Developers optimizing model routing
Route by task type and validation result. Keep a deeper fallback instead of assuming one model fits every request.
Cost context
Plan the cost of a useful result
Measure accepted cost
Compare the cost per record that passes validation, including retries and human review, rather than raw completion price.
Batch by stable shape
Stable inputs and schemas make throughput and error rates easier to monitor. Separate outliers instead of forcing one prompt to cover every case.
Keep escalation visible
Track how often Luna routes to a deeper model and whether those escalations improve the final result.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1/chat/completions' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "gpt-6-luna",
"messages": [
{
"role": "user",
"content": "What model are you"
}
]
}'FAQ
Frequently asked questions
What is GPT-6 Luna best suited for?
GPT-6 Luna is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is GPT-6 Luna priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call GPT-6 Luna?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
Is Luna just a shorter-context model?
The TokenHot catalog lists the same 1.05M context family. The practical difference is the task and cost trade-off; measure quality and correction rate rather than assuming context alone decides.
Can Luna process images?
The catalog lists text and image input. Verify the endpoint media shape and use human review for small text, handwriting or high-impact fields.
How should I route between Luna and Sol?
Use stable task classes, validation failures and representative fixtures. Escalate hard or uncertain work rather than silently accepting a weaker result.
Does a lower route price guarantee lower project cost?
No. Retries, schema failures and human correction matter. Compare the cost of an accepted result across the full workflow.
Get started
Build with GPT-6 Luna
API