| TiRex-238M / 82M | TiRex35M | Chronos-2120M | |
|---|---|---|---|
| Multivariate | yes | no | yes |
| Constant memory | yes | yes | no |
| Past covariates | yes | no | yes |
| Future covariates | yes | no | yes |
| Streaming | yes | no | no |
TiRex-238M / 82M parameters
- Multivariateyes
- Constant memoryyes
- Past covariatesyes
- Future covariatesyes
- Streamingyes
TiRex35M parameters
- Multivariateno
- Constant memoryyes
- Past covariatesno
- Future covariatesno
- Streamingno
Chronos-2120M parameters
- Multivariateyes
- Constant memoryno
- Past covariatesyes
- Future covariatesyes
- Streamingno
NeurIPS 2026Oral · top 0.4%
TiRex-2: Generalizing TiRex to Multivariate Data and Streaming
A recurrent, xLSTM-based time series foundation model for multivariate forecasting with past and future covariates. It keeps target variables strictly causal and streams at constant cost per patch.
@inproceedings{podest2026tirex2,
title = {TiRex-2: Generalizing TiRex to Multivariate Data and Streaming},
author = {Podest, Patrick and Pichler, Marco and B{\"u}rger, Elias and
Z{\'o}lyomi, Levente and Voggenberger, Bernhard and
Berghammer, Wilhelm and Klotz, Daniel and B{\"o}ck, Sebastian and
Klambauer, G{\"u}nter and Hochreiter, Sepp},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
year = {2026},
eprint = {2607.01204},
archivePrefix = {arXiv}
}