QuantumNous/new-api

What it solves

New API is an LLM gateway and AI asset management system designed to centralize the management of multiple AI model APIs. It solves the problem of managing fragmented API keys, handling diverse model formats, and tracking usage and costs across different providers (like OpenAI, Claude, Gemini, and others) from a single interface.

How it works

It acts as a proxy layer between the user and various AI model providers. It converts request formats between different APIs (e.g., converting OpenAI-compatible requests to Google Gemini or Claude Messages format) and provides a unified interface for authentication and routing. The system includes a management console for setting up channels, managing user quotas, and implementing intelligent routing (such as weighted random selection and automatic retries on failure).

Who it’s for

It is intended for organizations, developers, and authorized AI service providers who need to manage multi-model access, implement organization-level authentication, and track detailed usage analytics and cost accounting for internal or authorized enterprise customers.

Highlights

• Multi-Model Support: Supports OpenAI, Claude, Gemini, Azure, Midjourney-Proxy, and Suno-API, including specialized interfaces for chat, image, audio, video, and embeddings.
• Format Conversion: Automatically converts between OpenAI-compatible, Claude Messages, and Google Gemini formats.
• Usage Accounting: Provides internal top-up and quota allocation with support for EPay and Stripe, and per-request cost accounting.
• Intelligent Routing: Features channel weighted random routing, automatic failure retries, and user-level rate limiting.
• Enterprise Security: Supports OIDC unified authentication and authorization logins via Discord, Telegram, and LinuxDO.
• Reasoning Effort Control: Allows configuration of reasoning effort (low, medium, high) for OpenAI o3-mini and Gemini models, and thinking mode for Claude 3.7 Sonnet.