From the source
Users can compare models side-by-side by having them judge responses and voting on the best evaluation.
The platform includes 18 models and will produce a public leaderboard.
Early results show GPT-4 Turbo leading, but open-source models like Llama and Qwen are competitive.
