One API vs New API vs Sub2API: Why I Chose New API

Why Does Your AI Project Need an API Gateway?

When you’re juggling multiple AI services—OpenAI, Claude, Gemini, DeepSeek, and more—managing a pile of API keys, tracking usage, and dealing with different calling formats becomes an absolute nightmare. That’s where API Gateway platforms come in—they unify the chaotic AI ecosystem into one clean entry point.

Today I’m comparing three popular open-source solutions: One API, New API, and Sub2API, plus my real reasons for choosing New API.

Head-to-Head Comparison

One API — The Veteran Choice

One API is the pioneer in this space, developed by Song Quanpeng, and is the go-to choice for most people getting started.

Pros:

• Supports 50+ model providers—the widest ecosystem
• Richest community documentation—easy to find answers
• Battle-tested and stable in production
• Supports multi-server deployment and load balancing

Cons:

• Interface feels dated, hasn’t kept up with modern aesthetics
• No native Claude or Gemini API format support
• Missing support for some newer models (like o3, Gemini 2.0)
• New feature development is relatively slow

New API — The Rising Star

New API is essentially the spiritual successor to One API, led by Calcium-Ion. It retains the excellent architecture of One API while upgrading both features and user experience comprehensively.

Key Highlights:

1. Modern, Sleek UI

First impressions matter. New API’s interface design follows 2025 aesthetic trends—dark theme, smooth animations, clear information hierarchy. Using it is simply a pleasure.

2. Native Claude & Gemini Support

This is the most critical difference. New API supports Claude Messages API and Gemini API in their native formats, meaning you can use official Claude and Gemini SDKs directly without extra adaptation. For projects using newer models like gemini-2.0-flash-exp, this is a must-have.

3. Format Conversion Power

New API’s core selling point is cross-format conversion:

• OpenAI format ↔ Claude format
• OpenAI format → Gemini format
• Thinking-to-content support

This means you can call Claude or Gemini using OpenAI calling conventions, and vice versa. Change one line of code, switch your entire AI provider.

4. OpenAI Responses API & Realtime API

OpenAI’s Responses API (the foundation of Agent SDK) and Realtime API (for voice conversations) are currently only supported by New API. If you need to build Agent apps or real-time voice interactions, this is your only option.

5. Enterprise-Grade Billing

• Token usage tracking
• Cache billing support (Prompt Cache)
• Pay-per-use pricing strategies
• Multi-payment channel integration (Epay, Stripe)

6. Model Reasoning Effort Control

Control reasoning effort via suffixes:

• o3-mini-high — High reasoning effort
• gemini-2.5-flash-thinking — Enable chain-of-thought
• claude-3-7-sonnet-thinking — Claude thinking mode

Sub2API — Subscription Sharing Economy

Sub2API has a slightly different focus—it emphasizes subscription management and shared rides.

Key Features:

• Subscription-based API relay support (e.g., Claude Pro, GitHub Copilot)
• Built-in shared account (“拼车”) functionality
• Multi-channel payment: Epay, Alipay, WeChat Pay, Stripe
• Can aggregate multiple New API / One API sites

Best For:

• Running an AI API sharing/reselling service
• Sharing a subscription account among multiple users
• Integrating Chinese payment channels

Drawbacks:

• Smaller community, less complete documentation
• Essentially a secondary development on top of New API—loses some flexibility
• Overkill for pure personal use cases

My Choice: Why New API?

My final choice is New API, for practical reasons:

Reason 1: Gemini Image Generation Is My Must-Have

I want to use Gemini’s image generation capability (gemini-2.0-flash-preview-image-generation) to auto-generate article illustrations. This model requires Gemini native API format, which One API simply doesn’t support.

Reason 2: Format Conversion Is Too Good

New API’s OpenAI ↔ Claude ↔ Gemini format conversion lets me call different AI services with unified code. Use OpenAI today, switch to Claude tomorrow, try Gemini the day after—code barely changes.

Reason 3: Future-Proof

New API’s support for OpenAI Responses API and Realtime API means it can adapt to the latest AI SDKs and application scenarios. One API is feature-complete, but New API lets you use the newest and coolest stuff.

Reason 4: Dead Simple Deployment

# One command to launch
docker run --name new-api -d --restart always \
  -p 3000:3000 \
  -e TZ=Asia/Shanghai \
  -v ./data:/data \
  calciumion/new-api:latest

# Visit http://localhost:3000
# Default: root / 123456

Reason 5: Fully Compatible with One API

New API’s database is fully compatible with One API. If you already have One API data, you can migrate seamlessly. You’re never locked in—switch back to One API anytime.

Conclusion

If you’re like me—needing Gemini image generation, wanting to freely switch between OpenAI/Claude/Gemini, and wanting a beautiful interface—New API is the answer.

Resources

• One API GitHub
• New API GitHub
• Sub2API GitHub

Ready? Install New API and start your AI unified gateway journey.