全球节点,更低延迟
我们在美国、新加坡、韩国、香港、中国大陆均有自有节点,您在世界任何地方均可以低延迟、无障碍享受AI服务。
Use the most advanced models at the lowest rates. DeepRelay gives global developers a clean API route, transparent usage, and quick access from coding tools.
Streaming response ready, lower latency through Seoul relay, and token usage tracked in your dashboard.
Create API keys, track token usage, and inspect calls.
| Time | API Key | Model | Input | Output | Cost | Status |
|---|---|---|---|---|---|---|
No calls yet | ||||||
可多选,最多选择 3 个 Provider。
单选。模型数量会根据上一步选择的 Provider 实时筛选。
按 Provider 分组展示,最多选择 6 个模型。
| Order | Order number | Order amount | Status | Payment method | Created |
|---|---|---|---|---|---|
Loading orders... | |||||
Delete account
This removes your account and API keys from DeepRelay. This action cannot be undone.
Deeprelay支持您通过一个API调用文本、视频、音频、多模态等多种大模型。
OpenAI
面向复杂编程、Agent 协作和工具调用的新一代模型。
Qwen
统一理解文本、图像与视频内容,支持摘要、分析和视觉 Agent 工作流。
Google Gemini
最高支持文本、图片、视频、音频与 PDF 输入,适合长链路理解与执行。
OpenAI
通过专用图片 endpoint 完成文生图和图像编辑,与普通对话模型分类调用。
Google Gemini
通过原生音频调用路径输出自然语音,适合播客、有声内容与语音交互。
GLM
面向长上下文、深度推理、工具调用和结构化输出,适合高复杂度文本任务。
No. DeepRelay handles upstream model access and uses PayPal direct checkout for the current API credit route.
Not yet. Smart Plan is shown as a planned service layer. The active checkout route is API credit through PayPal direct.
Yes. Change the base URL to your DeepRelay endpoint and keep the same chat completion interface.
Recharge balance for token-based DeepRelay API calls.
| V4 Pro input | $2.00 / 1M |
| V4 Pro output | $3.80 / 1M |
| V4 Flash input | $0.20 / 1M |
| V4 Flash output | $0.35 / 1M |
| Status | Low-priority roadmap |
| Current launch | API pay-as-you-go first |
| Backend blocker | None for current launch |
Choose the providers and models your project needs, then create a scoped DeepRelay key in Dashboard.
Use the interactive workspace to select a provider and model, paste your key, and copy the exact prompt or command for your tool.
Open AI Agent SetupKeep protocol details, SDK examples, tool-specific configuration, and troubleshooting in the complete API documentation.
Open API documentationQuick Access guides the next step. AI Agent Setup generates personalized configuration. Documentation explains protocols, tools, and troubleshooting in detail.
Open a step-by-step guide with visual instructions, copy-ready configuration, verification, and troubleshooting.
Configure the Anthropic-compatible endpoint and verify your first request.
Add DeepRelay as a custom OpenAI provider and select an available model.
Set the custom provider, base URL, API key, and model in TRAE.
Write the provider configuration and run a minimal connection check.
Configure an authorized Codex model through the correct OpenAI route.
Choose the compatible protocol and connect DeepRelay without exposing your key.
Add a reusable DeepRelay provider and keep the key in an environment variable.
Create a custom model provider and test it inside your Dify workspace.
Add a custom provider, save the model, and confirm chat output.
Connect an Xcode AI extension to DeepRelay with copy-ready settings.
Set the provider environment variables and verify the active model.
Configure a compatible provider in the extension and test a coding request.
Add the DeepRelay assistant provider to Zed settings and validate it.
Point the CLI at DeepRelay and confirm the selected model responds.
Configure the OpenAI-compatible endpoint and run a safe smoke test.
Add DeepRelay credentials and verify the tool reads the intended model.
Use a framework-compatible provider configuration with a DeepRelay key.
Add an OpenAI connection and confirm models appear in the interface.
Initialize the SDK with DeepRelay and run a copy-ready Python example.
完整 Key 只展示一次,请立即复制并妥善保存。
Authorization: Bearer YOUR_KEY,base URL 为 https://deeprelayapi.cloud/v1。账户余额为 0 时需要先充值。This action cannot be undone. Any app using this key will stop working immediately.
DeepRelay creates a PayPal order for API credit and returns after capture.