
Independent company · Linz, Austria
NXAI
Linz company commercializing Sepp Hochreiter's xLSTM research, publishing xLSTM models, TiRex time-series forecasters and fast recurrent kernels.
Linz, Austria
- Founded
- 2024
- Type
- Independent company
- Based in
- Linz, Austria
- Key papers
- 4
- Featured projects
- 8
About
NXAI was set up in early 2024 in Linz, Austria, by Sepp Hochreiter, who leads the Institute for Machine Learning at JKU Linz, together with the company builder Netural X and PIERER Digital Holding. Hochreiter is Chief Scientist and Albert Ortig is CEO, according to press coverage.
The company aims to turn research on sequence modeling into industrial applications in engineering, robotics, automation and process control. It publishes the xLSTM architecture code, FlashRNN kernels, Vision-LSTM and the TiRex family of zero-shot time-series forecasting models.
Its GitHub organization dates from May 2024, and recent Hugging Face releases include TiRex-2 forecasting assets and KronLSTM language models.
Mission and ambitionsNXAI states that its mission is to turn pioneering research into market-relevant industrial AI by building scalable, generalist foundation models.
Known for
- 01
xLSTM recurrent language-model architecture
- 02
TiRex zero-shot time-series forecasting
- 03
FlashRNN kernels
- 04
Industry-focused AI from Linz
Milestones
- 2024
xLSTM repository
Reference implementation of the xLSTM architecture published in May 2024.
- 2024
Vision-LSTM
An xLSTM-based vision backbone repository created in June 2024.
- 2024
FlashRNN
Optimized kernels for recurrent networks released in November 2024.
- 2025
TiRex
A zero-shot time-series forecasting model built on xLSTM, with code from May 2025.
- 2026
TiRex-2
An updated forecasting model whose code was published in July 2026.
- 2026
KronLSTM models
1.5-billion-parameter recurrent language models published on Hugging Face in October 2026.
Open source
Official repository of the xLSTM (2,218 stars)
xLSTM as a generic vision backbone (490 stars)
Zero-shot forecasting model for long and short horizons (308 stars)
Fast RNN kernels with I/O awareness (190 stars)
- NX-AI/TiRexmodel
Zero-shot time-series forecasting model, the most downloaded NX-AI repository.
- NX-AI/TiRex-2model
Second generation of the TiRex forecaster.
- NX-AI/xLSTM-7bmodel
7B xLSTM language model weights.
- NX-AI/TiRex-demospace
Interactive demo of TiRex forecasting.
Research papers
Notable releases
Openness
The xLSTM code is Apache-2.0 and several models are public on Hugging Face.
People
- Sepp HochreiterCo-founder and Chief Scientist; head of the Institute for Machine Learning, JKU Linz
Inventor of the LSTM and senior author of the xLSTM and TiRex papers.
- Albert OrtigCEO
Leads the company, according to press coverage.
- Maximilian BeckLead author of xLSTM
First author of the xLSTM paper from the Hochreiter group.
- Andreas AuerLead author of TiRex
First author of the TiRex forecasting paper presented at NeurIPS 2025.
Listed from the lab's own pages and public profiles.
Contact and community
The company site prints a general contact address.
Sources
Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.



