The short version
Google's zero-shot foundation model for tabular prediction, following TimesFM.
From the HackerLinks archive
TabFM
Google's zero-shot foundation model for tabular prediction, following TimesFM.
At a glance:
First seen:2026-08-05
Last seen:2026-08-05
Times seen:1
Website:research.google
Why it caught our attention
A commenter judged it the strongest current approach for the tabular-prediction problem under discussion.
Where it surfaced on Hacker News
2026-08-05
“Google's TabFM [1] (and its previous TimesFM) seem the best approach so far on this area”
tough · recommendation
Direct HN comment
Why Large Language Models Fail at Tabular Prediction
Also surfaced in this discussion
TabPFN
Why Large Language Models Fail at Tabular Prediction
2026-08-05