GPT-6 Luna API
NewLLMThe GPT-6 Luna API is the fast, lower-priced GPT-6 route with 1M context, image input, and tool use.
- Input · USD / 1M tokens$0.2 USD
- Output · USD / 1M tokens$1.2 USD
- Context1M
- View tool pricing and billing rules
What are we making today?
Bring an idea, a question, or a first draft.
01
What this model is for
The GPT-6 Luna API is the fast, lower-priced GPT-6 route with 1M context, image input, and tool use.
GPT-6 Luna is OpenAI's low-priced tier of the GPT-6 family, released on September 22, 2026 as gpt-6-luna and built for focused, high-volume work. The catalog marks it as a fast tier and lists image input, tool use, a 1,050,000-token context window and the same 128,000-token output ceiling as GPT-6 Sol and GPT-6 Astra. Its knowledge cutoff is May 18, 2026, the latest in the family, and its input, output and cache-read rates are the lowest of the three.
Use Luna where a validator settles the answer cheaply: field extraction, label assignment, short rewrites and mechanical edits. The full effort range from none to max lets the same model skip reasoning for a clean classification or spend it on a messy page. GPT-6 Sol remains the choice for connected engineering decisions, and Claude Haiku 5.5 is the nearest alternative for code already on the Anthropic API.
Best-fit coding workloads
Support ticket classification
Assign a queue, priority and product area to each incoming ticket. Constrain the label set, include an unknown option, and run at none effort so a ticket that needs no reasoning does not spend reasoning tokens. Log every prediction with the field it relied on and review a sample of accepted labels each week.
Bulk rewriting at scale
Rewrite product descriptions, release notes or emails into a fixed style for thousands of records. Put the style rules and three examples in the system message, ask for plain text only, and check output against a banned-word list. A second pass for the failures stays inexpensive at Luna's listed rates.
Mechanical code edits
Apply a change you can describe exactly: rename a symbol, add a missing null check, convert a test to the new assertion style. Give the file, the rule and the test command, then run the suite on the patch. Send unresolved failures to GPT-6 Sol rather than retrying Luna many times.
02
Technical specification
Catalog facts and integration details verified for this route.
- Model ID
- gpt-6-luna
- Context
- 1M
- Inputs and tools
- Vision · Tool use · Fast responses
- API endpoint
- /v1/chat/completions
- Cache hit / 1M
- $0.02 USD
- Coding clients
- Codex · Cursor · Cline · Aider
Limits to plan around
- A well-formed JSON object or a plausible label can still be wrong, so every high-volume job needs a validator rather than a spot check.
- At max effort the reasoning tokens are billed as output and fill the window, so a long prompt leaves less room for the answer than the limits suggest.
- Function calling on Chat Completions works only with reasoning_effort set to none; reasoning and tools together require the Responses API.
- Output is text only. Image generation is a separate operation, and audio or video input is not listed for this model.
03
Compare nearby models
Compare workload fit and current listed token rates before choosing a route.
| Model | Best for | Context | Input | Output |
|---|---|---|---|---|
| GPT-6 LunaThis page gpt-6-luna | The default for bounded, high-volume steps that a cheap validator can check. | 1M | $0.2 USD Official price $0.1 USD | $1.2 USD Official price $0.5 USD |
| GPT-6 Sol gpt-6-sol | For connected engineering decisions where a wrong output costs far more than the tokens. | 1M | $0.225 USD Official price $2 USD | $1.35 USD Official price $10 USD |
| GPT-5.6 Luna gpt-5.6-luna | The older fast tier carries higher input, output and cache rates and a single price band. | 1M | $0.2 USD Official price $0.2 USD | $1.2 USD Official price $1.2 USD |
| Claude Haiku 5.5 claude-haiku-5-5 | The Anthropic alternative with a 1M window, for teams already on the Messages API. | 1M | $0.08 USD Official price $0.1 USD | $0.4 USD Official price $0.5 USD |
04
SDK quickstarts
Use one base URL and model ID from cURL, Python, or TypeScript.
curl https://api.omniakey.com/v1/chat/completions \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-6-luna",
"messages": [
{"role": "user", "content": "Review this pull request"}
],
"stream": true
}'05
Estimate token cost
Enter expected input and output tokens to compare current listed rates.
- OmniaKey estimate
- $0.044
- Official-rate estimate
- $0.02
- Estimated savings
- $0.00
Estimate covers input and output tokens only. Cache usage, retries, tools, and client-side token accounting can change the final charge.
06
Related coding guides
Deeper selection, setup, and troubleshooting guidance for this model family.
07
Sources and verification
Model facts and links were checked on 2026-10-27. Provider documentation remains the authority for upstream behavior.
Frequently asked questions
How much does GPT-6 Luna cost on OmniaKey?
OmniaKey currently charges $0.2 USD per 1M input tokens and $1.2 USD per 1M output tokens, in USD. Direct provider rates are separate and may change with input context length.
What is the context window for GPT-6 Luna?
The listed context window is 1M. Actual usable limits can also depend on the client and upstream request format.
What is GPT-6 Luna best used for?
Focused, repeatable work with a cheap check: classifying support tickets, extracting invoice fields, rewriting records into a fixed style and applying mechanical code edits. The catalog marks it as a fast tier and its listed rates are the lowest in the GPT-6 family, which suits high volumes. Keep a validator on every job, because a well-formed answer can still be wrong.
GPT-6 Luna or Claude Haiku 4.5: which should I pick?
Haiku 4.5 lists a 200K context window, while Luna lists 1,050,000 tokens, so a long repository or document set fits Luna without chunking. Haiku 4.5 comes from the older Anthropic generation, and Claude Haiku 5.5 is the closer comparison today. Choose by the API your code already speaks, then test both on your own tasks.
How large is the GPT-6 Luna model?
The catalog gives the context window rather than a parameter count: 1M tokens, 128,000 output tokens, image input and tool use. Treat those published limits, the effort range and the listed rates as the comparison points you can check, and decide from accepted outputs on your own tasks instead of a size figure.
How is GPT-6 Luna billed?
By input, output and cached-input token, with a cache-write charge on top. Reasoning tokens count as output, and a stable prefix keeps the cached rate in play. Batch and Flex are listed below Standard and Fast mode above it. Current rates are in the price table on the model page.
Can GPT-6 Luna use tools and images?
It accepts text and images and can call functions. Combine reasoning with tools through POST /v1/responses; on Chat Completions, function calling works only when reasoning_effort is none. Results come back as text, since the model does not return images. Use the id gpt-6-luna at https://api.omniakey.com/v1 and copy it exactly.