Skip to content

Latest commit

 

History

2 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

How fast do new AI models reach GitHub Copilot?

View the site →

Every time a frontier model ships, someone asks how long until it lands in GitHub Copilot. This measures it: the gap between a vendor's public release and the model showing up in Copilot, across 46 launches from September 2024 to August 2026.

The short answer

  • Median lag: 0 days. 70% of launches shipped in Copilot the same day the vendor announced them.
  • 80% shipped within a week.
  • Frontier models from Anthropic, OpenAI, and Google are effectively day-0 — GitHub is a launch partner.
  • Lag shows up on sub-variants (Codex-Max, mini tiers) and non-launch-partner vendors.
  • One model had negative lag: Grok Code Fast 1 was in Copilot as the stealth model "sonic" two days before xAI announced it.

Data

File What
data.json Canonical dataset, one record per launch
data.csv Same data, spreadsheet-friendly
build.py Source of truth. Edit ROWS, re-run, both files regenerate

Each record carries a vendorUrl and a changelogUrl so every number traces back to a public page.

python3 build.py   # regenerates data.json and data.csv
python3 -m http.server   # then open http://localhost:8000

No build step, no dependencies. The site is a single index.html that fetches data.json.

Method

  • Vendor date — the public announcement date from the vendor's own blog or newsroom. Not an API model-ID snapshot date; strings like claude-opus-4-5-20251101 are training cutoffs, not release dates.
  • Copilot date — the first GitHub Changelog entry announcing availability. Where a model shipped to public preview first, the preview date is used, since that is when users could actually select it.
  • Lag — calendar days between the two. Negative means Copilot was first.

Rows where the vendor date is genuinely ambiguous (experimental vs. GA, partner preview vs. public launch) carry a note field explaining the choice.

Corrections

Found a wrong date or a missing model? Open an issue or send a PR editing ROWS in build.py. Please include the vendor announcement URL and the changelog URL.

Disclaimer

Independent analysis of public data. Not affiliated with or endorsed by GitHub or any model vendor. Not a commitment, roadmap, or service-level agreement, and not a predictor of future releases.

License

MIT for the code. Data is compiled from public sources.

About

How fast do new AI models reach GitHub Copilot? 46 launches measured. Median lag: 0 days.

Topics

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages