Arena AI (formerly LMSYS Chatbot Arena) is the gold standard for AI model rankings. But it has no public API and no historical data.
So I built arena-ai-leaderboards — a GitHub repo that auto-fetches all 10 Arena AI leaderboards daily into structured JSON.
What It Does
A GitHub Actions workflow runs daily, auto-discovers all leaderboard categories from arena.ai, fetches full model rankings, parses them into structured JSON, validates against a schema, and commits to data/{YYYY-MM-DD}/.
No hardcoded category list — when Arena AI adds a new leaderboard, it gets picked up automatically.
10 Categories, 300+ Models
Text (LLM), Code, Vision, Text-to-Image, Text-to-Video, Image-to-Video, Image Edit, Document, Search, Video Edit.
Every model includes: rank, name, vendor, license type, ELO score, confidence interval, vote count.
Quick Access
REST API
Free, no auth. Hosted at api.wulong.dev:
# List all leaderboards
curl https://api.wulong.dev/arena-ai-leaderboards/v1/leaderboards
# Get LLM rankings
curl https://api.wulong.dev/arena-ai-leaderboards/v1/leaderboard?name=text
# Get a specific date
curl https://api.wulong.dev/arena-ai-leaderboards/v1/leaderboard?name=text-to-video\&date=2026-03-21
Raw GitHub JSON
curl https://raw.githubusercontent.com/oolong-tea-2026/arena-ai-leaderboards/main/data/latest.json
Python
import requests
text = requests.get(
"https://api.wulong.dev/arena-ai-leaderboards/v1/leaderboard?name=text"
).json()
for m in text["models"][:10]:
print(f"#{m["rank"]} {m["model"]} ({m["vendor"]}) — ELO {m["score"]}")
Why?
Arena AI is great for a quick glance but useless for trend analysis, automated monitoring, or research. No API, no historical data. This repo fixes that.
⭐ Star the repo if useful!