AA Models Tracker

Tracks the top 20 AI models from the Artificial Analysis LLM Leaderboard and publishes changes as an RSS feed.

How it works

• Scrapes the leaderboard page (parsed from Next.js RSC data)
• On first run: saves top 20 models to data/models.json and generates RSS entries
• On subsequent runs: diffs against stored data, writes new additions to data/feed.xml
• Each RSS entry includes the model description scraped from its individual page

Setup

poetry install

Run

poetry run models-track

Or:

poetry run python -m models_track.cli

Output

• data/models.json — current top 20 snapshot
• data/feed.xml — RSS feed with new model announcements

RSS entry format:

<item>
  <title>New model: GPT-5.4 from OpenAI, intel index 57.17</title>
  <link>https://artificialanalysis.ai/models/gpt-5-4</link>
  <description>GPT-5.4 is amongst the leading models in intelligence...</description>
</item>

Automation

GitHub Actions runs every 6 hours (cron 0 */6 * * *) and can be triggered manually from the Actions tab.

Development

poetry run ruff check .
poetry run ruff format --check .
poetry run mypy src/