# NX-AI/xlstm

Official repository of the xLSTM.

Repository: https://github.com/NX-AI/xlstm
Canonical: https://ross.abutalabs.com/products/xlstm
Homepage: https://www.nx-ai.com/
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, deep-learning-architecture, llm, machine-learning, nlp, rnn
Last push: 2026-08-26T14:48:07+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 8, longevity 60
- inputs: {"age_days": 847, "days_push": 7, "days_rel": 462, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2198, forks 187 (observed 2026-08-28T04:06:25.170687+00:00)

## What it is
Official PyTorch implementation of xLSTM, an extended Long Short-Term Memory recurrent architecture with exponential gating and matrix memory, including the xLSTM 7B large language model. It serves as a research library and model codebase for training and running recurrent language models competitive with Transformers and State Space Models.

## Use cases
- train a recurrent language model as an alternative to transformers
- run the xLSTM 7B model for fast and efficient inference
- experiment with new RNN architectures for sequence modeling
- implement exponential gating and matrix memory in PyTorch
- compare xLSTM against state space models on language modeling benchmarks
- fine-tune an xLSTM model on custom text data

## When to choose
- you need an efficient recurrent alternative to Transformers for language modeling
- you want fast inference with constant-memory recurrent state
- you are researching modern RNN/LSTM architectures
- you want to use or extend the official xLSTM implementation from NXAI

## When to avoid
- you need a mature ecosystem with broad tooling like Hugging Face Transformers
- your workload depends on Transformer-specific tooling or fine-tuning pipelines
- you need non-PyTorch support
- you only need a pretrained off-the-shelf chat model with minimal setup

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, llm-training
- domain: deep-learning, large-language-models, machine-learning
- platform: python, cross-platform
- tags: rnn, lstm, sequence-modeling, pytorch, language-modeling, research, transformer-alternative, natural-language-processing, gpu, linux

## Member repositories
- NX-AI/xlstm (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:25.170687+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:47:08.547643+00:00, confidence not recorded.
  - readme: https://github.com/NX-AI/xlstm (fetched 2026-08-28T04:06:25.170687+00:00, sha 4de04912d03a)
  - homepage: https://www.nx-ai.com/ (fetched 2026-08-29T10:27:22.709017+00:00, sha 8dcd9feae118)
  - registry_pypi: https://pypi.org/pypi/xlstm/json (fetched 2026-08-29T10:27:22.719620+00:00, sha 083ad7284a57)
  - site_page: https://www.nx-ai.com/en/company (fetched 2026-08-29T10:27:22.717887+00:00, sha 8afb9e895358)
- Data as of 2026-08-30T08:39:29.467469+00:00.