EfficientCurrent model
GPT-5.6 Luna
gpt-5.6-lunaCost-sensitive, high-concurrency, and high-volume text and coding work.
Viewing nowGPT-5.6 Luna targets cost-sensitive, high-throughput workloads such as bulk classification, extraction, rewriting, and lightweight coding. HiAPI currently exposes text, multi-turn, and function tool calling through the streaming Responses API.
Provider: OpenAI
Category: text generation
Endpoint: /v1/responses
Status: Available
Cost: --
GPT-5.6 Luna targets cost-sensitive, high-throughput workloads such as bulk classification, extraction, rewriting, and lightweight coding. HiAPI currently exposes text, multi-turn, and function tool calling through the streaming Responses API.
API access
Endpoint
POST /v1/responses
Base URL
https://api.hiapi.ai
Uses the OpenAI Responses format with streaming required. One HiAPI key works across all available models.
Input price
2,000 Credits
/ 1M tokens
Output price
12,000 Credits
/ 1M tokens
Context
1.05M
context window
Max output
128K
output tokens
Ask a question to stream the answer and inspect token usage, latency, and estimated cost.
Call GPT-5.6 Luna with the streaming OpenAI Responses format. Copy a ready-to-use cURL, Python, or Node.js example below.
/v1/responsescurl -N -X POST "https://api.hiapi.ai/v1/responses" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.6-luna",
"input": [
{
"type": "message",
"role": "user",
"content": [
{
"type": "input_text",
"text": "把下面的用户反馈分成产品、支付、性能和其他四类，并分别提取一句摘要。"
}
]
}
],
"stream": true,
"store": false,
"reasoning": {
"effort": "medium"
}
}'Tip: Replace YOUR_API_KEY with your actual API key from the API Keys page.
Uses the OpenAI Responses format. Keep stream: true, send input as a message array, and set reasoning effort with reasoning.effort.
Responses API
Start with core fields, then enable advanced capabilities as needed.
The fields that form a valid request
Enable reasoning, context, and tools as needed
Max output is a model specification limit, not a per-request generation target. Examples show only the runtime fields currently available and verified on HiAPI.
Billing is based on actual input, output, and cached token usage. All prices below are shown per 1 million tokens.
Prompts and context sent to the model
Responses and reasoning generated by the model
These specifications reflect the current public HiAPI text request contract. Undisclosed capabilities are not inferred by the page.
GPT-5.6 Luna targets cost-sensitive, high-throughput workloads such as bulk classification, extraction, rewriting, and lightweight coding. HiAPI currently exposes text, multi-turn, and function tool calling through the streaming Responses API.
HiAPI exposes this model through /v1/responses. Test prompts and parameters in the Playground, then use the same HiAPI API key in your server application.
The Playground and API share the same model ID, so three steps take a tested prompt into production.
Step 1
Test instructions, input, and reasoning.effort in the Playground.
Step 2
View an existing key or create another after signing in; one key works across all available models.
Step 3
Send requests to /v1/responses and track cost with the usage object.
GPT-5.6 Luna is the efficient GPT-5.6 tier for cost-sensitive, high-concurrency, or high-volume text and coding workloads. HiAPI currently exposes text, multi-turn conversations, and function tool calls through the streaming Responses API with one HiAPI API key.
Choose Sol for quality-first complex professional work, Terra for a balanced mix of capability, latency, and cost, or Luna for cost-sensitive and high-throughput workloads. Evaluate all three with the same representative prompts, then compare answer quality, time to first token, total latency, and actual token cost.
Send POST requests to /v1/responses, set model to gpt-5.6-luna, pass an input message array with stream=true and store=false, and authenticate with Authorization: Bearer HIAPI_API_KEY. This is not Chat Completions, so do not use messages or choices as request and response fields.
Set reasoning.effort to none, low, medium, high, xhigh, or max; the default is medium when omitted. Use none/low for latency-sensitive work, medium as a general starting point, high/xhigh for quality-first tasks, and max only for the hardest requests that justify extra latency and token use.
This endpoint requires stream=true and emits typed Responses API SSE events. Read text from response.output_text.delta, completion from response.completed, and failures from error. Do not reuse the Chat Completions choices[0].delta parser.
The examples use store=false, so the most reliable approach is to keep conversation state in your application and replay the required user, assistant, and tool items in the next input. Use previous_response_id only with a compatible storage and response-chaining strategy. Prior tokens brought back into context are still billed.
Yes. Declare functions in tools and control selection with tool_choice. After a function_call, run the local tool and return function_call_output with the same call_id before continuing. This page lists only the tool capabilities verified on HiAPI; unlisted hosted tools are not guaranteed.
OpenAI currently documents a 1,050,000-token context window, 922,000 maximum input tokens, and 128,000 maximum output tokens. Input, output, and reasoning all consume context. The displayed maximum is a model specification; use the API examples and capability parameters on this page as the runtime contract.
HiAPI bills actual input, output, and cached-input tokens separately, with reasoning tokens counted as output. The model and pricing pages show current public rates; the Responses API usage object and HiAPI usage logs record final usage and charges.
Set the SDK base_url to https://api.hiapi.ai/v1, use a HiAPI API key, call client.responses.create, replace messages with input, and set model to gpt-5.6-luna. HiAPI requests must keep stream=true and consume typed SSE events.