@upstash/model-multiplexer

Eliminate 429 Rate Limit Errors Forever 🚀

A lightweight, zero-dependency TypeScript library that combines the quotas of multiple LLM providers into a single unified API. Never hit rate limits again by automatically distributing your requests across OpenAI, Claude, Gemini, and other providers.

The Problem: Rate Limits Kill Your App

• ❌ Error 429: "Rate limit exceeded" stops your application
• ❌ Quota exhaustion: Single provider limits constrain your throughput
• ❌ Unpredictable failures: Rate limits hit at the worst possible moments
• ❌ Manual failover: Switching providers requires code changes

The Solution: Combined Quotas

✅ 10x Higher Throughput: Combine OpenAI + Claude + Gemini quotas
✅ Zero 429 Errors: Automatic failover when one provider hits limits
✅ Seamless Integration: Drop-in replacement for OpenAI SDK
✅ Smart Load Balancing: Weight-based distribution across providers

Key Benefits

• 🚀 Quota Multiplication: Combine rate limits from multiple providers for massive throughput
• 🛡️ 429 Error Elimination: Automatic failover prevents rate limit failures
• ⚡ Zero Downtime: Seamless switching between providers when limits hit
• 🔌 OpenAI Compatible: Works with existing OpenAI SDK code
• 🎯 Zero Dependencies: Lightweight with no runtime dependencies
• 📊 Usage Analytics: Track which providers are hitting limits

Installation

npm install @upstash/model-multiplexer openai

Note: You need to install openai as it's a peer dependency

Quick Start

import { Multiplexer } from "@upstash/model-multiplexer";
import OpenAI from "openai";

// Create client instances
const claude = new OpenAI({
  apiKey: process.env.ANTHROPIC_API_KEY,
  baseURL: "https://api.anthropic.com/v1/",
});

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: "https://api.openai.com/v1",
});

// Initialize multiplexer
const multiplexer = new Multiplexer();

// Add models with weights and specific model names
multiplexer.addModel(claude, 5, "claude-sonnet-4-0");
multiplexer.addModel(openai, 3, "gpt-4.1-mini");

// Use like a regular OpenAI client
const completion = await multiplexer.chat.completions.create({
  model: "claude-sonnet-4-0", // Will be overridden by selected model
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "What is the capital of France?" },
  ],
});

console.log(completion.choices[0].message.content);

Multi-Provider Setup

import { Multiplexer } from "@upstash/model-multiplexer";
import OpenAI from "openai";

// Set up clients for different providers
const claude = new OpenAI({
  apiKey: process.env.ANTHROPIC_API_KEY,
  baseURL: "https://api.anthropic.com/v1/",
});

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: "https://api.openai.com/v1",
});

const gemini = new OpenAI({
  apiKey: process.env.GEMINI_API_KEY,
  baseURL: "https://generativelanguage.googleapis.com/v1beta/",
});

const multiplexer = new Multiplexer();

// Add primary models (higher quality, potentially stricter rate limits)
multiplexer.addModel(claude, 5, "claude-sonnet-4-0");
multiplexer.addModel(claude, 3, "claude-opus-4-0"); // Same provider, separate quota!
multiplexer.addModel(gemini, 4, "gemini-2.5-pro-preview-05-06");

// Add fallback models (cheaper, higher availability)
multiplexer.addFallbackModel(openai, 5, "gpt-4.1-mini");
multiplexer.addFallbackModel(openai, 3, "gpt-4.1"); // Same provider, separate quota!
multiplexer.addFallbackModel(gemini, 3, "gemini-2.0-flash");

// Result: Combined quotas from multiple models + multiple providers = massive throughput

API Reference

Creating a Multiplexer

const multiplexer = new Multiplexer();

Adding Models

// Add a primary model
multiplexer.addModel(client: OpenAI, weight: number, modelName: string)

// Add a fallback model
multiplexer.addFallbackModel(client: OpenAI, weight: number, modelName: string)

Parameters:

• client: OpenAI-compatible client instance
• weight: Positive integer for weight-based selection (higher = more likely to be selected)
• modelName: Specific model name to use (e.g., "gpt-4.1-mini", "claude-sonnet-4-0")

Getting Statistics

const stats = multiplexer.getStats();
// Returns: Record<string, { success: number; rateLimited: number; failed: number }>

Resetting the Multiplexer

multiplexer.reset(); // Clears all models and resets state

Streaming Support

const stream = (await multiplexer.chat.completions.create({
  model: "claude-sonnet-4-0",
  messages: [{ role: "user", content: "Write a poem about AI." }],
  stream: true,
})) as AsyncIterable<OpenAI.Chat.Completions.ChatCompletionChunk>;

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content || "");
}

How Quota Combining Works

Single Model:        [GPT-4: 10,000 RPM] ❌ 429 Error at 10,001 requests
Multiple Providers:  [OpenAI: 10K] + [Claude: 15K] + [Gemini: 20K] = 45,000 RPM ✅
Multiple Models:     [GPT-4: 10K] + [GPT-4-mini: 50K] + [Claude: 15K] = 75,000 RPM ✅✅

The Magic Behind Zero 429 Errors

1. Quota Multiplication: Your effective rate limit becomes the SUM of all models (even from same provider)
2. Isolated Model Limits: Each model has separate rate limits (GPT-4 + GPT-4-mini = 2x OpenAI quota)
3. Smart Distribution: Requests are distributed across all models based on weights
4. Instant Failover: When Model A hits 429, traffic instantly routes to Model B
5. Cross-Provider Redundancy: Combine models from multiple providers for maximum resilience
6. Transparent Operation: Your code sees one unified API, not multiple models/providers

Real-World Impact

Single Model Approach:

• 1,000 requests/minute → ❌ 429 error when GPT-4 limit hit

Multi-Model Same Provider:

• 1,000 requests/minute → ✅ distributed as 400 (GPT-4) + 600 (GPT-4-mini) → success

Multi-Provider Setup:

• 1,000 requests/minute → ✅ distributed as 300 (GPT-4) + 300 (GPT-4-mini) + 200 (Claude) + 200 (Gemini) → maximum resilience

Environment Variables

Set up your API keys:

export OPENAI_API_KEY="your-openai-key"
export ANTHROPIC_API_KEY="your-anthropic-key"
export GEMINI_API_KEY="your-gemini-key"

Examples

Check out the examples directory for more detailed usage patterns.

TypeScript Support

Full TypeScript support with proper type definitions included.

import { Multiplexer } from "@upstash/model-multiplexer";
// All OpenAI types are available through the peer dependency

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT

About Upstash

Upstash provides serverless databases and messaging infrastructure for modern applications.