Arena AI Model ELO Tracker

AI researchers and product teams want visibility into how flagship models perform over time. This dashboard visualizes historical ELO ratings from Arena AI, letting users track if models degrade post-launch or improve with updates.

LLM benchmarking anxiety is real right now — teams have watched GPT-4 visibly degrade post-launch and the "did they lobotomize it?" discourse recurs every few months, so demand for longitudinal performance tracking is genuine. Chatbot Arena (LMSYS) is the closest reference point, but it doesn't offer historical trend visualization or exportable time-series data, which is the actual gap here. The $500–2k/mo ceiling makes sense for a narrow analytics dashboard, but it also reveals the ceiling — this is a feature, not a product, and a well-resourced team at Weights & Biases, Scale AI, or even Hugging Face could ship it in a sprint. The biggest risk is that Arena's underlying data is public and scraped by default, meaning anyone can replicate the core value prop trivially, which collapses willingness to pay and makes it hard to defend beyond being first to build a clean UI.