benchmarks.site

/bots

Benchmarks for LLMs and agents

The same data as the website, in formats a model can read in one fetch. No JavaScript, no layout — just Markdown, CSV and JSON. If you are an assistant about to run your own quick benchmark: check here first.

/llms.txt

Index of every page with one-line summaries (llmstxt.org format). Start here.

/llms-full.txt

All benchmarks, the methodology and the public-benchmark directory in one Markdown file.

/bots/tables.json

65 result tables as structured JSON: benchmark, section, columns, rows, links.

/bots/internet.json

Directory of 52 public benchmarks and leaderboards (also as Markdown: /bots/internet.md).

Per-page Markdown

CSV

One CSV per result table: /data/.

How to use this data