foundational prediction

Size

What a deployment carries.

1 MB10 MB100 MB1.0 GBlaplace (pure Python)108 KBlaplace++ (weekly, monthly)180 KB 1.7× laplacelaplace (Rust core)2 MB 15× laplaceAutoARIMA2 MB 17× laplaceAutoETS2 MB 17× laplaceAutoTheta2 MB 17× laplaceGARCH-t4 MB 40× laplaceTimestral-28M111 MB 1,000× laplaceTiRex141 MB 1,300× laplaceChronos-Bolt-small191 MB 1,800× laplaceTimeH2O-v1435 MB 4,000× laplaceTimestral-209M837 MB 7,700× laplaceTimesFM 2.5925 MB 8,500× laplacet0-gamma1.0 GB 9,400× laplaceTimesFM 3.04.0 GB 37,000× laplace

Bar length is the fourth root of the size, between a log and a linear scale.

modelsizemeasured as
laplace (pure Python)108 KBpublished wheel, no dependencies
laplace++ (weekly, monthly)180 KBpublished wheel plus the 4,202-parameter checkpoint; runs on PyTorch, not counted
laplace (Rust core)2 MBpublished wheel plus the compiled core
AutoARIMA2 MBthe installed statsforecast package, dependencies not counted
AutoETS2 MBthe installed statsforecast package, dependencies not counted
AutoTheta2 MBthe installed statsforecast package, dependencies not counted
GARCH-t4 MBthe installed arch package, dependencies not counted
Timestral-28M111 MBweight files as cached from the Hub
TiRex141 MBweight files as cached from the Hub
Chronos-Bolt-small191 MBweight files as cached from the Hub
TimeH2O-v1435 MBweight files as cached from the Hub
Timestral-209M837 MBweight files as cached from the Hub
TimesFM 2.5925 MBweight files as cached from the Hub
t0-gamma1,022 MBweight files as cached from the Hub
TimesFM 3.03,969 MBweight files as cached from the Hub

Built 2026-11-09.