本页演示通过 知能API 兼容代理请求 gpt-5.6-sol 的推理摘要。reasoning.summary 返回的是模型生成的摘要,不是完整、逐字的内部思维链。
gpt-5.6-sol
reasoning.summary
https://api.apexrouter.top/v1/responses
API 密钥安全
示例令牌为不可用占位符。请妥善保管真实 知能API 令牌,不要在公开仓库或日志中泄露。
none
low
medium
high
xhigh
max
auto
# ============================================================ # GPT-5 深度思考模式调用示例(SDK 版) # 功能:通过 OpenAI Python SDK 调用 Responses API # ============================================================ from openai import OpenAI # ---------- 初始化客户端 ---------- client = OpenAI( api_key="sk-*****************************", base_url="https://api.apexrouter.top/v1" ) # ---------- 发送请求 ---------- response = client.responses.create( model="gpt-5.6-sol", # 使用的模型 input="一加到十最复杂的方法", # 用户输入的问题 reasoning={ "effort": "low", # 思考等级 "summary": "auto" # 思考摘要:auto 自动返回思考过程 } ) # ---------- 输出结果 ---------- # 遍历 output 列表,分别提取「思考内容」和「响应结果」 for item in response.output: if item.type == "reasoning": print("【思考内容】") for content in item.summary: print(content.text) print() elif item.type == "message": print("【响应结果】") for content in item.content: print(content.text)
# ============================================================ # GPT-5 深度思考模式调用示例 # 功能:调用 OpenAI Responses API,获取模型的思考过程与最终回答 # ============================================================ import requests import json import sys import io # ---------- 终端编码设置 ---------- # 解决 Windows 终端输出中文乱码问题 sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8') # ---------- API 配置 ---------- url = "https://api.apexrouter.top/v1/responses" api_key = "sk-*****************************" headers = { "Content-Type": "application/json", "Authorization": f"Bearer {api_key}" } # ---------- 请求参数 ---------- data = { "model": "gpt-5.6-sol", # 使用的模型 "input": "一加到十最复杂的方法", # 用户输入的问题 "reasoning": { "effort": "low", # 思考等级 "summary": "auto" # 思考摘要:auto 自动返回思考过程 } } # ---------- 发送请求 ----------· response = requests.post(url, headers=headers, json=data) result = response.json() # ---------- 解析并输出结果 ---------- # 遍历 output 列表,分别提取「思考内容」和「响应结果」 for item in result.get("output", []): if item.get("type") == "reasoning": print("【思考内容】") for content in item.get("summary", []): print(content.get("text", "")) print() elif item.get("type") == "message": print("【响应结果】") for content in item.get("content", []): print(content.get("text", ""))
curl -X POST https://api.apexrouter.top/v1/responses \ -H "Content-Type: application/json" \ -H "Authorization: sk-***************************************" \ -d '{ "model": "gpt-5.6-sol", "input": "比较归并排序和快速排序,并说明各自更适合的场景。", "reasoning": {"effort": "low", "summary": "auto"} }'
{ "id": "resp_example_gpt56_reasoning_001", "object": "response", "status": "completed", "model": "gpt-5.6-sol", "output": [ { "id": "rs_example_gpt56_reasoning_001", "type": "reasoning", "summary": [ { "type": "summary_text", "text": "比较两种算法的时间复杂度、稳定性、额外空间与典型使用场景。" } ] }, { "id": "msg_example_gpt56_reasoning_001", "type": "message", "status": "completed", "role": "assistant", "content": [ { "type": "output_text", "text": "归并排序稳定且最坏时间复杂度为 O(n log n),适合外部排序和要求稳定性的场景;快速排序通常原地且平均性能优秀,适合内存中的通用排序。", "annotations": [] } ] } ] }
© 2026 知能API GPT-5.6 Sol 推理摘要