# New API

> An open-source next-generation LLM gateway and AI asset management system that unifies multiple AI providers under OpenAI-compatible, Claude-compatible, or Gemini-compatible interfaces.

New API is an open-source LLM gateway and AI asset management system built by QuantumNous, licensed under AGPLv3. It enables developers and organizations to aggregate multiple AI model providers behind a single unified interface, supporting self-hosted private deployment via Docker. The project is actively maintained on GitHub and describes itself as a next-generation fork of the One API project.

## What It Is

New API is a centralized AI gateway written in Go that sits between your applications and upstream AI providers. It cross-converts various LLMs into OpenAI-compatible, Claude-compatible, or Gemini-compatible API formats, letting teams standardize on one interface regardless of which underlying model they use. It is designed for lawful, authorized API gateway scenarios, organization-level authentication, multi-model management, usage analytics, cost accounting, and private deployment.

## Format Conversion and Model Support

A core capability of New API is bidirectional format translation across major API schemas:

- **OpenAI Compatible ⇄ Claude Messages** — full bidirectional conversion
- **OpenAI Compatible → Google Gemini** — outbound conversion
- **Google Gemini → OpenAI Compatible** — text only; function calling not yet supported
- **OpenAI Responses format** — supported as both input and output
- **Realtime API** — OpenAI Realtime including Azure
- **Rerank models** — Cohere and Jina
- **Midjourney-Proxy, Suno API, Dify ChatFlow** — additional model type support

Reasoning effort configuration is supported for OpenAI o-series models, Claude thinking models, and Google Gemini 2.5 series, including fine-grained thinking budget control.

## Intelligent Routing and Operations

New API provides channel-level management features suited for multi-team or enterprise internal deployments:

- Weighted random channel selection
- Automatic retry on upstream failure
- User-level model rate limiting
- Token grouping and model restrictions
- Per-request, usage-based, and cache-hit cost accounting
- Cache billing statistics for OpenAI, Azure, DeepSeek, Claude, Qwen, and other supported models
- Internal top-up and quota allocation with EPay and Stripe integrations

## Deployment Model

The project is designed for self-hosted private deployment. The recommended path is Docker Compose, with Docker command-line and BaoTa Panel one-click install also supported. Database options include SQLite (default, requires `/data` mount), MySQL ≥ 5.7.8, and PostgreSQL ≥ 9.6. Redis is supported for caching and is required for multi-machine deployments alongside `SESSION_SECRET` and `CRYPTO_SECRET` environment variables. The Docker image is published as `calciumion/new-api:latest`.

The project homepage includes an official statement warning that New API has never publicly sold API access and recommends developers pull source code from the official GitHub repository for private local deployment.

## Update: v1.0.0-rc.10

The latest release as of the available sources is **v1.0.0-rc.10**, published on 2026-05-26. The repository was last pushed to on 2026-06-10, indicating active development. The project originated as a fork of One API (MIT License) and has accumulated over 38,000 GitHub stars and 8,700 forks according to the repository metadata. The project is featured on Product Hunt and listed on Trendshift and HelloGitHub. QuantumNous notes that the project is multilingual, supporting Simplified Chinese, Traditional Chinese, English, French, and Japanese interfaces.

## Features
- Unified LLM gateway for multiple AI providers
- OpenAI-compatible API interface
- Claude Messages format support
- Google Gemini format support
- Bidirectional format conversion (OpenAI ⇄ Claude)
- OpenAI Realtime API support (including Azure)
- Rerank model support (Cohere, Jina)
- Midjourney-Proxy integration
- Suno API integration
- Dify ChatFlow support
- Weighted random channel routing
- Automatic retry on upstream failure
- User-level model rate limiting
- Token grouping and model restrictions
- Per-request and usage-based cost accounting
- Cache billing statistics
- Internal quota allocation and top-up (EPay, Stripe)
- Discord, LinuxDO, Telegram, OIDC authorization login
- Visual data dashboard and analytics
- Multi-language UI (Chinese, English, French, Japanese)
- Docker and Docker Compose deployment
- SQLite, MySQL, PostgreSQL database support
- Redis caching support
- Reasoning effort configuration for OpenAI, Claude, Gemini
- Thinking-to-content functionality

## Integrations
OpenAI, Azure OpenAI, Anthropic Claude, Google Gemini, DeepSeek, Qwen (Alibaba), Midjourney (via Midjourney-Proxy), Suno API, Dify, Cohere, Jina, EPay, Stripe, Discord (OAuth), Telegram (OAuth), LinuxDO (OAuth), Redis, MySQL, PostgreSQL, Docker, Pyroscope

## Platforms
LINUX, WEB, API, JETBRAINS_PLUGIN, CLI

## Pricing
Open Source

## Version
v1.0.0-rc.10

## Links
- Website: https://www.newapi.ai
- Documentation: https://www.newapi.ai/en/docs
- Repository: https://github.com/QuantumNous/new-api
- EveryDev.ai: https://www.everydev.ai/tools/new-api