Google's 200M-parameter time-series foundation model with 16k context
by codepawl on 3/31/2026, 5:21:59 AM
https://github.com/google-research/timesfm
Comments
by: EmilStenstrom
I somehow find the concept of a general time series model strange. How can the same model predict egg prices in Italy, and global inflation in a reliable way?<p>And how would you even use this model, given that there are no explanations that help you trust where the prediction comes from…
3/31/2026, 5:45:35 AM
by: EmilStenstrom
Here is the link to the blogpost, that actually describe what this is: <a href="https://github.com/google-research/timesfm?tab=readme-ov-file" rel="nofollow">https://github.com/google-research/timesfm?tab=readme-ov-fil...</a>
3/31/2026, 5:41:17 AM
by: kuu
It would be nice to add (2024) to the title, this is not news (see: <a href="https://research.google/blog/a-decoder-only-foundation-model-for-time-series-forecasting/" rel="nofollow">https://research.google/blog/a-decoder-only-foundation-model...</a>)
3/31/2026, 7:39:04 AM
by: dash2
So the time series are provided with no context? It's just trained on lots of sets of numbers? Then you give it a new set of numbers and it guesses the rest, again with no context?<p>My guess as to how this would work: the machine will first guess from the data alone if this is one of the categories it has already seen/inferred (share prices, google trend cat searches etc.) Then it'll output a plausible completion for the category.<p>That doesn't seem as if it will work well for any categories outside the training data. I would rather just use either a simple model (ARIMA or whatever) or a theoretically-informed model. But what do I know.
3/31/2026, 6:27:11 AM
by: wiradikusuma
Also: <a href="https://github.com/Nixtla/nixtla" rel="nofollow">https://github.com/Nixtla/nixtla</a> and <a href="https://facebook.github.io/prophet/" rel="nofollow">https://facebook.github.io/prophet/</a>
3/31/2026, 6:07:30 AM
by: ra
This has been around a few months now, has anyone built anything on it?
3/31/2026, 6:26:29 AM
by: croemer
(2024)
3/31/2026, 8:07:02 AM
by: Foobar8568
Somehow I missed that one. Are there any competition on this?<p>I always had difficulties with ML and time series, I'll need to try that out.
3/31/2026, 5:36:18 AM
by: raghavMultilipi
This has been around a few months now, has anyone built anything on it?
3/31/2026, 7:10:46 AM
by: emsign
Can this finally break the stock markets?
3/31/2026, 7:20:02 AM
by: jdthedisciple
Let me be blunt: Shannon would tell us that time forecasting is bullshit:<p>There is infinitely more entropy in the real world out there than any model can even remotely capture.<p>The world is not minecraft.
3/31/2026, 6:26:44 AM
by: charlotte12345
[dead]
3/31/2026, 7:15:00 AM