Skip to main content
For example, for the text-to-speech model tts-1-hd from OpenAI, you can not send more than 7 requests in minute. Any extra request automatically fails. There are many real-world use cases where it’s possible to run into rate limits:
  • When your requests have very high input-tokens count or a very long context, you can hit token thresholds
  • When you are running a complex and long prompts pipeline that fires hundreds of requests at once, you can hit both token & request limits

​
Here’s an overview of rate limits imposed by various providers:

LLM ProviderExample ModelRate Limits
OpenAIgpt-5Tier 1:500 Requests per Minute10,000 Tokens per Minute10,000 Requests per Day
AnthropicAll modelsTier 1:50 RPM50,000 TPM1 Million Tokens per Day
CohereCo.Generate modelsProduction Key:10,000 RPM
AnyscaleAll modelsEndpoints:30 concurrent requests
Perplexity AImixtral-8x7b-instruct24 RPM16,000 TPM
Together AIAll modelsPaid:100 RPM
Generally, developers tackle getting rate limiting with a few tricks: caching common responses, queuing the requests, reducing the number of requests sent, etc. Using Portkey, you can solve this by just adding a few lines of code to your app once. In this cookbook, we’ll show you 2 ways of ensuring that your app never gets rate limited again:

​
1. Install Portkey SDK

npm i portkey-ai
Let’s start by making a generic chat.completions call using the Portkey SDK: 
import Portkey from "portkey-ai";

const portkey = new Portkey({
    apiKey: process.env.PORTKEY_API_KEY
});

response = await portkey.chat.completions.create({
    messages: [{role: "user", content: "Hello!"}],
    model: "@openai-prod/gpt-5-mini"  // @provider-slug/model-name
});

console.log(response.choices[0].message.content);
Add your OpenAI credentials in Model Catalog to get a provider slug like @openai-prod. To ensure your request doesn’t get rate limited, we’ll utilise Portkey’s fallback & loadbalance features:

​
2. Fallback to Alternative LLMs

With Portkey, you can write a call routing strategy that helps you fallback from one provider to another provider in case of rate limit errors. This is done by passing a Config object while instantiating your Portkey client: 
import Portkey from "portkey-ai";

const portkey = new Portkey({
    apiKey: process.env.PORTKEY_API_KEY,
    config: JSON.stringify({
        strategy: {mode: "fallback", on_status_codes: [429]},
        targets: [
            {override_params: {model: "@openai-prod/gpt-5-mini"}},
            {override_params: {model: "@anthropic-prod/claude-4.5-sonnet", max_tokens: 254}}
        ]
    })
});

​
In this Config object,

  • The routing strategy is set as fallback
  • on_status_codes param ensures that the fallback is only triggered on the 429 error code, which is generated for rate limit errors
  • targets array contains the model slugs from Model Catalog and the order of the fallback
  • The override_params in the second target lets you add more params for the specific provider. (max_tokens for Anthropic in this case)
That’s it! The rest of your code remains the same. Based on the Config, Portkey will do the orchestration and ensure that Anthropic is called as a fallback option whenever you get rate limit errors on OpenAI. Fallback is still reactive - it only gets triggered once you actually get an error. Portkey also lets you tackle rate limiting proactively:

​
3. Load Balance Among Multiple LLMs

Instead of sending all your requests to a single provider on a single account, you can split your traffic across multiple provider accounts using Portkey - this ensures that a single account does not get overburdened with requests and thus avoids rate limits. It is very easy to setup this “loadbalancing” using Portkey - just write the relevant loadbalance Config and pass it while instantiating your Portkey client once: 
import Portkey from "portkey-ai";

const portkey = new Portkey({
    apiKey: process.env.PORTKEY_API_KEY,
    config: JSON.stringify({
        strategy: {mode: "loadbalance"},
        targets: [
            {override_params: {model: "@openai-account1/gpt-5-mini"}, weight: 1},
            {override_params: {model: "@openai-account2/gpt-5-mini"}, weight: 1},
            {override_params: {model: "@openai-account3/gpt-5-mini"}, weight: 1}
        ]
    })
});

​
In this Config object:

  • The routing strategy is set as loadbalance
  • targets contain 3 different OpenAI provider slugs from Model Catalog (representing 3 different accounts), all with equal weight - which means Portkey will split the traffic 1/3rd equally among the 3
With Loadbalance on different accounts, you can effectively multiply your rate limits and make your app much more robust.

​
That’s it!

You can read more on fallbacks here, and on loadbalance here. If you want a deep dive on how Configs work on Portkey, check out the docs here.
Last modified on February 4, 2026