leaderboard

name	leaderboard
version	1.3.0
description	Track AI model leaderboard rankings over time and detect rank/score changes between snapshots

Leaderboard Snapshot Tracker

Track AI model leaderboard rankings over time using SQLite snapshots. Detect new models, removed models, rank changes, and score changes between dates.

Supported Leaderboards

Leaderboard	URL	Modality
LMSYS Chatbot Arena	https://lmsys.org	Text, Vision
LMArena	https://lmarena.ai	Text, Vision
HuggingFace Open LLM	https://huggingface.co/spaces/open-llm-leaderboard	Text
Artificial Analysis	https://artificialanalysis.ai	Text, Image, Video
Scale AI SEAL	https://scale.com/leaderboard	Text

Usage

Save a snapshot

cd {SKILL_DIR} && python3 skills/leaderboard/scripts/leaderboard_snapshot.py save \
  --leaderboard "chatbot-arena" \
  --date 2026-04-14 \
  --data '[{"model": "claude-4-opus", "rank": 1, "score": 1350}]'

Prints the diff against the previous snapshot (new models, rank changes, score changes).

View latest snapshot

cd {SKILL_DIR} && python3 skills/leaderboard/scripts/leaderboard_snapshot.py latest \
  --leaderboard "chatbot-arena"

Data Storage

Snapshots are stored in ~/.cache/morning-ai/leaderboard.db (SQLite). Each entry has:

• leaderboard — leaderboard identifier
• model — model name
• rank — position on the leaderboard
• score — numeric score (ELO, accuracy, etc.)
• snapshot_date — date of the snapshot

Integration with Main Workflow

This skill is currently a standalone utility. It can be integrated into the main morning-ai workflow as a Benchmark data source:

1. As a collector: Scrape leaderboard pages → save snapshot → diff against previous → generate TrackerItem entries for rank changes
2. As a report section: Add a "Leaderboard Movement" section to the daily report showing rank/score deltas

To integrate, a collector module (lib/leaderboard_collector.py) would:

• Fetch current leaderboard data from supported sites
• Call save_snapshot() to persist
• Call diff_snapshot() to detect changes
• Convert significant changes (new #1 model, big rank jumps) into TrackerItem objects with Benchmark type