NA

New API Integration Guide

This guide will help you integrate GPT-Load proxy service with New API platform, supporting OpenAI, Gemini, Gemini OpenAI compatible, and Anthropic channel types.

Prerequisites

Ensure you have successfully deployed and started the GPT-Load service, running on http://localhost:3001 by default

1
OpenAI Channel Configuration

Configuration Steps

Select Type

Select "OpenAI" type in New API

Configure Key

sk-123456

Use the proxy key you configured in GPT-Load

Set API Address

http://localhost:3001/proxy/openai

Where "openai" is the group name you configured in GPT-Load

Add Models and Test

Add model list and perform testing verification

New API OpenAI Configuration Screenshot

Click image to enlarge

2
Gemini Channel Configuration

Configuration Steps

Select Type

Select "Google Gemini" type in New API

Configure Key

sk-123456

Use the proxy key you configured in GPT-Load

Set API Address

http://localhost:3001/proxy/gemini

Where "gemini" is the group name you configured in GPT-Load

Model List Configuration

Due to a bug in New API's Gemini type model list retrieval, manual model name entry is recommended when creating new channels

New API Gemini Configuration Screenshot

Click image to enlarge

Model List Retrieval Notice

New API's Gemini type may not automatically retrieve model lists when creating new channels. It's recommended to manually enter model names first. After creation, model lists can usually be retrieved normally in edit mode.

Gemini Thinking Configuration

Thinking Feature Setup

Enable Thinking Adaptation

In New API's System Settings → Model Related Settings, select "Enable Gemini Thinking Suffix Adaptation"

Use Thinking Suffix

• -thinking：Enable thinking mode
• -nothinking：Disable thinking mode

Note: gemini-2.5-pro model cannot disable thinking functionality

New API Gemini Thinking Configuration Screenshot

Click image to enlarge

Important Warning

After configuring thinking functionality in New API, do not configure thinking-related parameters again in GPT-Load's parameter override to avoid conflicts from duplicate configuration.

Gemini Search Feature Configuration

Enable Official Search Tools

Configuration Choice

Search models can be configured in either New API or GPT-Load (only configure in one place)

Parameter Configuration

Add the following configuration to parameter override:

{
  "tools": [
    {
      "google_search": {}
    }
  ]
}

After configuration, the official search tool functionality will be enabled

3
Gemini OpenAI Compatible Format

Configuration Steps

Select Type

Select "Custom Channel" type in New API

Configure Key

sk-123456

Use the proxy key you configured in GPT-Load

Set Full Address

http://localhost:3001/proxy/gemini/v1beta/openai/chat/completions

Where "gemini" is the group name you configured in GPT-Load

Manually Add Models

Compatible format cannot automatically retrieve model lists, manual model name entry is required

New API Gemini OpenAI Compatible Format Configuration Screenshot

Click image to enlarge

Compatible Mode Limitations

When using OpenAI compatible mode, thinking and search parameters cannot be configured. For these advanced features, use native Gemini type instead.

4
Anthropic (Claude) Channel Configuration

Configuration Steps

Select Type

Select "Anthropic Claude" type in New API

Configure Key

sk-123456

Use the proxy key you configured in GPT-Load

Set API Address

http://localhost:3001/proxy/anthropic

Where "anthropic" is the group name you configured in GPT-Load

New API Anthropic Configuration Screenshot

Click image to enlarge

Important Notes

Configuration Notes

• Replace the GPT-Load access address in examples with your actual service address

• Group names in paths (like openai, gemini, anthropic) must match your actual configuration in GPT-Load

• Test configuration in a small scope first, then use officially after confirmation

Related Resources

New API Official Documentation
GPT-Load Configuration Guide
Channel Types Guide