1/5🚀 Introducing TiRex-2— our next-generation time series foundation model. Time series forecasting in the real world is streaming: -new observations continuously arrive, -variables interact, -some covariates are known for the future, -models must update predictions efficiently.
Users highlight TiRex-2's efficiency and SOTA benchmark results via xLSTM as a strong recurrent alternative to Transformers for time series forecasting.
No Digg Deeper questions have been answered for this story yet.
4/5 On two leading zero-shot benchmarks, GIFT-Eval and fev-bench, TiRex-2 achieves state-of-the-art performance while requiring substantially fewer active parameters than competing foundation models.
3/5 Most foundation models are Transformers, requiring repeated processing of the history. Not with TiRex-2. With xLSTM, TiRex-2 combines: ✅ Multivariate forecasting ✅Past & future-known covariates ✅Constant-cost streaming inference ✅State-of-the-art zero-shot forecasting
This work demonstrates that recurrent, memory-centric architectures remain a powerful alternative to Transformers—not only for language, but also for large-scale time series forecasting.
4/5 On two leading zero-shot benchmarks, GIFT-Eval and fev-bench, TiRex-2 achieves state-of-the-art performance while requiring substantially fewer active parameters than competing foundation models.
2/5 Paper and Model: Paper: https://arxiv.org/html/2607.01204v1 Huggingface: https://huggingface.co/NX-AI/TiRex-2 Github: https://github.com/NX-AI/tirex-2
1/5🚀 Introducing TiRex-2— our next-generation time series foundation model. Time series forecasting in the real world is streaming: -new observations continuously arrive, -variables interact, -some covariates are known for the future, -models must update predictions efficiently.
3/5 Most foundation models are Transformers, requiring repeated processing of the history. Not with TiRex-2. With xLSTM, TiRex-2 combines: ✅ Multivariate forecasting ✅Past & future-known covariates ✅Constant-cost streaming inference ✅State-of-the-art zero-shot forecasting
2/5 Paper and Model: Paper: https://arxiv.org/html/2607.01204v1 Huggingface: https://huggingface.co/NX-AI/TiRex-2 Github: https://github.com/NX-AI/tirex-2
