How much better has general Glean Assistant gotten over the past couple of months?

anthonyga

I've noticed that the Glean Assistant has been steadily improving over time.

Personally, I've seen significant enhancements in both the accuracy and overall quality of its responses compared to late last year. This improvement may be partly due to the recent transition to GPT-4.1.

It would be beneficial if Glean shared updates about new features and improvements being rolled out for the Glean Assistant. Communicating these changes could help rebuild trust among users who may have had less positive experiences in the past.

neill

Seems to have improved considerably due to both 4.1 and to the more agentic retrieval of results.

I ran a hallucination benchmark that was specific to our data. It went from hallucinating in 7 of 20 to hallucinating in 2 of 20. I reran the test 5 times for each prompt to ensure consistency.