Perplexity API OpenAI Compatibility: What Is Supported

Developers building AI-powered applications often ask whether the Perplexity API works with tools and libraries built for OpenAI. The Perplexity API is designed to be compatible with the OpenAI client libraries and API format, making migration or integration simpler. This article explains exactly which OpenAI features are supported by the Perplexity API and which are not. You will learn the key differences in endpoints, parameters, authentication, and streaming behavior.

Perplexity API Design and OpenAI Compatibility Overview

The Perplexity API is built on the same REST architecture as OpenAI’s API. This means you can use the same HTTP methods, headers, and JSON body structure for most requests. The goal is to let developers switch between providers with minimal code changes.

The main difference is the base URL and model names. Perplexity does not support every OpenAI feature. Some parameters like function_call and tools are missing. The API is focused on chat completions and does not offer image generation, embeddings, or fine-tuning.

Base URL and Endpoint Path

OpenAI’s chat completions endpoint is at https://api.openai.com/v1/chat/completions. Perplexity uses https://api.perplexity.ai/chat/completions. The path after the base URL is identical. This means you only need to change the base URL and authentication token to switch from OpenAI to Perplexity.

Authentication Method

Both APIs authenticate using an API key passed as a Bearer token in the Authorization header. For example: Authorization: Bearer pplx-xxxxxxxxxxxxxxxx. Perplexity API keys start with pplx-. You generate them from the Perplexity account dashboard under API Keys.

Model Names

OpenAI uses model names like gpt-4, gpt-3.5-turbo, and gpt-4-turbo. Perplexity uses its own model names such as sonar-pro, sonar-small, and sonar-medium. You must replace the model name in your request body. The full list of available models is at https://api.perplexity.ai/models.

Supported OpenAI Features in the Perplexity API

The following table lists the OpenAI features that Perplexity supports. If a feature is not listed, assume it is not supported.

Feature	OpenAI Parameter	Perplexity Support
Chat completions	/v1/chat/completions	Yes
Streaming	stream: true	Yes
System messages	role: system	Yes
User messages	role: user	Yes
Assistant messages	role: assistant	Yes
Temperature	temperature	Yes
Max tokens	max_tokens	Yes
Top P	top_p	Yes
Stop sequences	stop	Yes
Presence penalty	presence_penalty	Yes
Frequency penalty	frequency_penalty	Yes
N (number of choices)	n	Yes
Logprobs	logprobs	No
Functions	functions	No
Tools	tools	No
Tool choice	tool_choice	No
Response format (JSON mode)	response_format	No

Steps to Switch an OpenAI Application to Perplexity

Follow these steps to migrate a simple chat application from OpenAI to Perplexity. These steps assume you are using the official OpenAI Python library or a similar client.

1. Get your Perplexity API key
   Log in to perplexity.ai/settings/api. Click Generate API Key. Copy the key that starts with pplx-.
2. Change the base URL in your client
   In the OpenAI client, set base_url to https://api.perplexity.ai. For example: client = OpenAI(api_key='pplx-...', base_url='https://api.perplexity.ai').
3. Replace the model name
   Change the model parameter from gpt-4 to sonar-pro or another Perplexity model. Example: model='sonar-pro'.
4. Remove unsupported parameters
   Delete any functions, tools, tool_choice, logprobs, or response_format fields from your request body.
5. Test the connection
   Send a simple chat completion request. If you receive a 401 error, check your API key. If you receive a 404 error, verify the endpoint path.

Common Misconceptions and Limitations

Can I Use the OpenAI Python Library Directly?

Yes, you can use the openai Python library with the Perplexity API by changing the base URL and API key. The library will format requests correctly. However, the library may try to call unsupported endpoints like /v1/models or /v1/embeddings. You need to handle those calls separately or disable them.

Does Perplexity Support Streaming Responses?

Yes, Perplexity supports streaming with stream: true. The response format is identical to OpenAI’s streaming format. Each chunk contains a choices[0].delta object with content or role. You can use the same streaming logic.

Are Function Calls or Tool Use Available?

No, Perplexity does not support function calls or tool use. The functions, tools, and tool_choice parameters are ignored or cause an error. If your application relies on function calling, you need to implement a separate mechanism or choose a different provider.

Does Perplexity Support Embeddings or Image Generation?

No, Perplexity only provides chat completions. It does not offer embeddings, image generation, or audio processing. Use OpenAI or another provider for those tasks.

Perplexity API vs OpenAI API: Key Differences

Item	Perplexity API	OpenAI API
Base URL	https://api.perplexity.ai	https://api.openai.com
Chat completions endpoint	/chat/completions	/v1/chat/completions
Model naming	sonar-pro, sonar-small, sonar-medium	gpt-4, gpt-3.5-turbo, etc
Function calling	Not supported	Supported
Tool use	Not supported	Supported
Streaming	Supported	Supported
Logprobs	Not supported	Supported
JSON mode	Not supported	Supported
Embeddings	Not available	Available
Image generation	Not available	Available

The Perplexity API offers a compatible interface for basic chat completions with streaming and standard sampling parameters. It does not support advanced OpenAI features like function calling, tools, JSON mode, or embeddings. To migrate an existing application, change the base URL to https://api.perplexity.ai, replace the model name with a Perplexity model, and remove any unsupported parameters. For a deeper integration, consider using the Perplexity-specific features like the search_domain parameter for web search or academic search results. Test your application thoroughly after switching because some edge cases like error response formats may differ.