Perplexity AI offers a unique approach to language models with their fresh llms. These online llms provide real-time, up-to-date information by combining search engine power with traditional language models. This approach sets them apart from other open models, allowing for more accurate and timely responses. The API, accessible through a pro subscription, is still in early beta but fully functional. The single chat completions API, compatible with open AI, is a powerful feature. And with a $5 monthly credit for pro subscribers, it’s worth a try. The ability to invoke models like Lama 270B and online llms opens up a world of possibilities for knowledge-based chatbots and applications. So buckle up and stay tuned for the latest updates on generative AI from perplexity Labs. 👀
Overview
Welcome back to part two of the Perplexity AI Series. In this walkthrough, we’ll explore the developer experience of using the platform. Perplexity AI offers an API endpoint for developers which exposes models for inference. This demo will walk you through the API exposed by Perplexity and invoking the Llama 2 model. If you have a pro subscription, you’ll receive a $5 credit every month to try the API.
Developer Experience
The API is evolving and still in the early beta stage, but it’s fully functional. One standout feature is that it’s open AI compatible, specifically designed for the chat completion API. This provides a seamless experience for users familiar with open AI’s endpoints.
Supported Models:
- Cod Lama
- Lama 270B chat
- Mistal 7B instruct
- Mixl 8X 7B instruct
Unique Approach
Perplexity has its own set of foundation models, the 7B chat, and the 70B chat. These models leverage online llms or fresh llms, capable of delivering up-to-date, real-time responses. These fresh llms are built based on a novel approach summarized in a recent paper published on November 22nd, 2023.
Fresh Prompt Approach
The approach leverages a combination of traditional llms with search engines to deliver contextually relevant and up-to-date information. This innovative approach is named "Fresh LLMS: Refreshing LLMS with Search Engine Augmentation".
Walkthrough Demo
The demo will involve setting the model to Lama 270B, followed by a question prompt to demonstrate the capabilities of the API. Running the same prompt through both the Lama 270B and Perplexity’s 7B online model will showcase the discrepancy in results and response time.
API Usage
This API provides powerful possibilities for integrating with various applications. Building chat bots, combining logical reasoning with online searches, or creating rag applications are all achievable using the Perplexity API.
Choosing the Right Model
Perplexity offers two online models – 7B and 70B. While both models share a common context window of 4096 tokens, understanding the nuances between the models is crucial for selecting the most suitable option.
Next Steps
The Perplexity API & its potential use cases will be further explored in upcoming videos and articles. Stay tuned for more updates and insights from the world of generative AI.
Key Takeaways:
- Perplexity AI API offers an open AI compatible chat completion endpoint.
- The API boasts support for various models with unique capabilities.
- Fresh LLMS leverage a novel approach to deliver real-time responses.
Thank you for watching! Don’t forget to subscribe for more updates from the world of generative AI.
EN_GB