Weekly digest
The week's releases, once a week
A Monday email of what the LLM Releases changelog recorded in the seven days before it was sent.
LLM Releases is an independent, sourced record of large language model launches, updates, deprecations, and retirements — frontier and open-weight. The digest is that record, delivered once a week, so you do not have to check the catalog every day. It is a release digest, not a general AI news roundup and not a leaderboard.
What is in the email
The message is generated from the changelog. Nothing extra is written for it. The subject line is “LLM Releases — this week”, plus how many entries are included. The opening line says how many updates were tracked in the last seven days.
Each entry has:
- the changelog title, linked to that model's page on this site
- the lab name, when the catalog has one
- the event date
The list is every changelog entry in that window, not a hand-picked subset. That includes releases, updates, deprecations, and retirements, and also other tracked changes such as price changes and benchmark updates. The email does not add a benchmark table, a separate summary, or the primary-source URL. Those stay on the changelog and the model page.
If nothing was tracked in those seven days, no email is sent.
For an on-site view, the release radar is a separate 30-day briefing, and latest releases is the newest models in the catalog. Neither one is the digest itself.
Cadence
The send is scheduled for Mondays at 15:00 UTC. It covers the changelog for the seven days before it runs. When email delivery is configured, messages go out through Resend, one address per message.
How to subscribe
Enter your email below. Each digest includes an unsubscribe link for that address. How the address is stored is described in the privacy notice.
Prefer not to share an email address? Follow without email via the RSS feed. The feed carries a short summary and a source link when the catalog has them. The email is the shorter list described above.
Why this digest exists
The field ships something new almost every week. Official lab posts are the primary sources, and they are easy to miss if you follow more than one lab. This digest stays narrow on purpose: what the changelog recorded, with a link back to the model page.
The project is not affiliated with any model developer. Where labs are named, it is to attribute their work. Read about the project and the methodology for how entries are checked and logged.
If something is wrong
Spotted an error, a missing model, or a benchmark that needs a source? Email corrections@llm-releases.com with a primary-source link. Corrections are versioned, so fixes are visible in the record.