Benchmark Radar — a live database and search engine for all AI benchmarks

🔎 Benchmark Radar — a live search engine for AI benchmarks

The usual story with benchmarks: there's an article with numbers, but the datasets, code, and settings behind the published scores have to be hunted down piece by piece. Benchmark Radar was built for exactly this task — a "live" database and search engine for AI benchmarks. The system's goal is to let you find those very datasets, code, and settings behind published scores.

The key difference from a static catalog is updating. The system daily collects and indexes benchmark articles, repositories, datasets, and releases .🔄 Inside — a searchable catalog of benchmarks plus tracking of mentions in model cards.

By evaluation categories, the coverage is as follows: LLM evaluation, agentic and tool-use benchmarks, code, reasoning, safety, and domain tests.

The tool is intended for benchmark researchers and developers of LLMs and other AI systems. The material is dated September 14, 2026 📅, published on HuggingFace Papers under ID 2609.11115 — huggingface.co/papers/2609.11115