Check the Body, Not Only the Status
Errors such as 401 invalid API key, 402 insufficient credits or 429 rate limited can come back with HTTP status 200, so read the response body before treating a call as successful.
Automatically append previous messages to maintain multi-turn context. May increase token usage.
Function calling
Web Access
Reasoning effort
Stream Response
How can I help you today?
Automatically append previous messages to maintain multi-turn context. May increase token usage.
Function calling
Web Access
Reasoning effort
Stream Response
How can I help you today?
README
Complete guide to using GPT-6 Sol and Luna
GPT-6 Luna API and GPT-6 Sol API on Kie.ai
GPT-6 Luna and GPT-6 Sol are OpenAI GPT-6 models trained with methods similar to GPT-6 Astra, which OpenAI describes as bringing Astra's advances to faster, more affordable models. GPT-6 Sol has since been upgraded to GPT-6.1 Sol. On Kie.ai, both GPT-6 Luna and GPT-6 Sol are available as API models.
New to GPT-6 Sol? Read what it is and how it compares →
GPT-6 Luna and GPT-6 Sol API Parameters on Kie.ai
GPT-6 Luna and GPT-6 Sol take the same parameters on Kie.ai; only the model name changes.
| Parameter | Value |
|---|---|
| Models | gpt-6-luna, gpt-6-sol |
| Endpoint | POST /codex/v1/responses, compatible with the OpenAI Responses API |
| Context window | 1,050,000 tokens |
| Maximum input | 922,000 tokens |
| Maximum output | 128,000 tokens |
| Input types | Text, images and files, mixed in one message |
| Reasoning effort | low, medium, high, xhigh or max; default medium |
| Tools | Web search or function calling, not both in one request |
| Streaming | On by default (server-sent events); set stream to false for a single JSON response |
| Pricing | Shown in the pricing card above; select a model to see its price |
GPT-6 Luna vs GPT-6 Sol: When Luna Is Enough
OpenAI calls GPT-6 Luna its most efficient model for focused, high-volume tasks, while GPT-6 Sol is built to take on difficult work tasks. On OpenAI's DeepSWE v1.1 coding benchmark at maximum reasoning effort, GPT-6 Sol scores 68.8% and GPT-6 Luna 66.6%, and LMArena's text leaderboard gives GPT-6 Sol 1456 and GPT-6 Luna 1443 (as of October 8, 2026). On Kie.ai, Luna costs less than Sol. A practical default is Luna for high-volume, focused requests and Sol for harder work tasks; GPT-6 Sol has since been upgraded to GPT-6.1 Sol, and OpenAI points to GPT-6 Astra for the most demanding work.
| GPT-6 Luna | GPT-6 Sol | |
|---|---|---|
| How OpenAI positions it | Its most efficient model for focused, high-volume tasks | Built to take on difficult work tasks |
| DeepSWE v1.1, max reasoning effort (OpenAI-reported) | 66.6% | 68.8% |
| LMArena text score (Oct 8, 2026) | 1443 | 1456 |
| Artificial Analysis Intelligence Index (Oct 2026) | 38 | Not listed |
| Newer version | — | Upgraded to GPT-6.1 Sol |
| Available on Kie.ai | Yes | Yes |
What OpenAI Reports for GPT-6 Luna and GPT-6 Sol
OpenAI trained GPT-6 Sol and Luna with methods similar to GPT-6 Astra and reports these improvements over their GPT-5.6 predecessors.
Business Workflows Across Apps
On AutomationBench, a test of business workflows across apps, GPT-6 Luna at high reasoning effort scores 5.4 points above its predecessor.
Fewer Factual Mistakes
On OpenAI's factuality evaluation, built from real conversations where users flagged mistakes, GPT-6 Sol makes about half as many mistakes as its predecessor, and GPT-6 Luna also improves.
Code Ready to Merge
On FrontierCode, which checks whether coding agents produce changes ready to merge into real codebases, GPT-6 Sol improves substantially over GPT-5.6 Sol.
Clearer, Shorter Replies
OpenAI brought GPT-6 Astra's communication style to Sol and Luna: more clarity, less jargon, fewer low-value details and slightly shorter answers.
Typical Uses for GPT-6 Luna and GPT-6 Sol
These are the kinds of work each model fits, based on how OpenAI positions them.
Focused Requests at Volume
Well-defined requests that run many times a day, where Luna's efficiency matters most.
Coding in Real Codebases
Writing and fixing code in existing repositories, with Sol for the harder changes.
Business Workflows Across Apps
Multi-step tasks that move information between apps.
Professional Work With Agents
Complex professional workflows run by agents, where Sol takes on the difficult steps.
How to Use the GPT-6 Luna and GPT-6 Sol APIs on Kie.ai
Step 1. Get Your API Key
Step 2. Test Your Prompt in the Playground
Step 3. Send a Request
Tips Before You Ship
Web Search or Function Calling, Not Both
One request can use web search or your own function tools, not both. When you pass function tools, set tool_choice to auto; without function tools, leave tool_choice out.
Why Use GPT-6 Luna and GPT-6 Sol on Kie.ai
Same Key Across the Kie.ai Market
All models in the Kie.ai market use the same authentication method: an Authorization: Bearer YOUR_API_KEY header with the key from your Kie.ai API Key Management page.
Free Credits to Start
New Kie.ai users receive 80 free credits for testing. Your account has one credit balance that all generation services draw from, and you can check it with the Get Remaining Credits API.
Nothing to Host or Install
Kie.ai provides this model as an API with a playground; there is no app to download or install. There are no GPUs or model weights to manage.
FAQs About GPT-6 Sol API and GPT-6 Luna API
Pricing is shown in the pricing card above. Select GPT-6 Luna or GPT-6 Sol on the card to see each model's price.
Not on Kie.ai: each request uses credits, as shown in the pricing card above. New Kie.ai users receive 80 free credits for testing.
OpenAI positions Luna for focused, high-volume tasks and Sol for difficult work tasks. Sol scores higher on OpenAI's DeepSWE results and on LMArena, and Luna costs less on Kie.ai; the comparison table above has the scores.
As of October 2026, OpenAI's GPT-6.1 release page covers only GPT-6.1 Sol, and the Luna model available on Kie.ai is GPT-6 Luna (gpt-6-luna).
GPT-6 Luna is the GPT-6 model OpenAI describes as its most efficient, built for focused, high-volume tasks. On Kie.ai it is the API model gpt-6-luna; its parameters, results and uses are covered above.
OpenAI lists a 1,050,000-token context window, with up to 922,000 input tokens and 128,000 output tokens. The parameter table above has the other limits.
Create a Kie.ai API key, then send a request to /codex/v1/responses with model set to gpt-6-luna. The steps above show the full flow.
Yes. On Kie.ai you can still call gpt-6-sol, although OpenAI has since upgraded it to GPT-6.1 Sol.