Many-Shot Predictions

Recent findings have shown that we can achieve greater predictability in language model outputs by running the same prompt multiple times.

This app provides an easy way to run the same prompt N times and visualize the results, helping you understand the distribution and consistency of model responses.

You will need to clone the GitHub repository and add your own API keys for each platform (OpenAI, Anthropic, Gemini, etc.). If you prefer not to use those platforms, you can try a local browser model, but it is slower.