360Brew was a LinkedIn® research model—not the name of the platform’s live feed algorithm. That distinction matters because many articles and posts turned an interesting technical experiment into confident advice about reach, profiles, posting frequency, and content structure that the available evidence did not support.
Key Takeaways
- 360Brew was a 150-billion-parameter research and pre-production model for ranking and recommendation tasks.
- LinkedIn® tested it with a small group of members and later shut down the test.
- LinkedIn® has explicitly said 360Brew is not part of its ranking system.
- The research paper does not justify claims that 360Brew controls reach or requires creators to change their profiles and posts.
- Useful LinkedIn® strategy should be based on confirmed platform information, your own evidence, and the needs of the audience.
What 360Brew Actually Was
The 360Brew research paper describes a decoder-only foundation model developed to explore whether a single model with a textual interface could perform numerous ranking and recommendation tasks.
The authors described 360Brew V1.0 as a 150-billion-parameter pre-production model trained and fine-tuned using LinkedIn® data and tasks. Its reported results concerned offline metrics across more than 30 predictive tasks. That is not the same as confirming a platform-wide production deployment.
What LinkedIn® Later Confirmed
In April 2026, LinkedIn® VP of Engineering Tim Jurka answered the central question directly. He said 360Brew had been tested with a small group of members, was judged not to be the right fit for the platform, and the test was shut down.
His LinkedIn® clarification states that 360Brew is not used as part of the platform’s ranking system.
Why the Confusion Spread
The paper described an ambitious model capable of interpreting textual representations of members, content, behaviors, and social connections. Commentators then converted those research capabilities into claims about what every user should do to gain reach.
Those claims commonly included:
- 360Brew had replaced LinkedIn®’s previous ranking systems;
- the model was reading every profile and post like a person;
- profile-to-post consistency had become a new ranking requirement;
- specific content pillars, entities, formats, or posting patterns would train the model; and
- recent changes in reach could be attributed directly to 360Brew.
LinkedIn®’s clarification means those claims should not be presented as explanations of the live feed.
What the Research Can—and Cannot—Tell Us
The paper remains useful as research. It shows how LinkedIn® explored the use of a large language model across multiple ranking and recommendation tasks and why a textual interface could reduce dependence on task-specific feature engineering.
It does not provide a creator playbook. It does not establish that changing a headline, repeating chosen topics, naming more entities, increasing dwell time, or generating early engagement will satisfy 360Brew. It also does not prove that the model caused an individual account’s rise or fall in distribution.
How to Approach LinkedIn® Reach Instead
Do not diagnose performance from a named-algorithm story. Begin with evidence you can actually examine:
- compare a meaningful group of similar posts rather than isolated examples;
- review changes in subjects, formats, cadence, audience, and participation;
- assess who saw and responded to the content, not only the total reach;
- check whether the profile, Company Page, and destination pages support the message;
- separate official platform information from independent observation and speculation; and
- test one meaningful change at a time.
Clear positioning, credible expertise, useful content, and genuine professional participation remain sensible practices because they help people understand and trust the work—not because 360Brew requires them.
The Practical Conclusion
360Brew should be discussed as a discontinued LinkedIn® experiment with an informative research paper behind it. It should not be used as the explanation for current reach, a label for LinkedIn®’s entire algorithm, or the foundation for selling supposedly algorithm-proof tactics.
For a broader method of evaluating platform claims, read LinkedIn® Algorithm Changes: How to Respond Without Guessing.
