Not Diamond is an AI model evaluation and recommendation framework that helps you predictively determine which LLM is best-suited to respond to each input in your application, improving accuracy by up to 25% while lowering costs up to 10x.

Key features

Train your own custom routers: Leverage any evaluation data across any set of models and any range of inputs to train your own custom routers optimized to your use case.
Maximize output quality: Not Diamond outperforms every individual foundation model on major evaluation benchmarks by predictively recommending the best model for every input.
Reduce cost and latency: Define explicit cost and latency tradeoffs to efficiently leverage smaller and cheaper models without degrading quality.
RAG auto-evaluation and optimization: Generate test data from your document store, auto-optimize your retrieval parameters, and evaluate various LLMs on your RAG pipeline.
Client-side or gateway model requests: Receive model recommendations and then make your requests client-side or through our gateway. Available through our API or through VPC deployments.
Python, TypeScript, and REST API support: Easily integrate Not Diamond across a variety of stacks.

Getting started

Making your first API request with Not Diamond takes less than 5 minutes. To get started:

Create an account at app.notdiamond.ai
Create a Not Diamond API key
Jump into the quickstart example

Alternatively, you can join one of 50,000 weekly active users of our Not Diamond-powered chatbot to see what routing feels like as an end-user. We also have a Not Diamond-powered RAG app that you can use to ask any questions you have about Not Diamond.