Instella

by AMD

About the model:

Open LLM trained entirely on AMD GPUs.

Model type:

Text

Model performance class:

Full

Link to the model:

https://huggingface.co/amd/Instella-3B-Long-Instruct

Base models:

Instella-3B

End model:

Instella-3B-Instruct

End model license:

Research-only RAIL-M5

About the organisation:

AMD, a major chip manufacturer.

Link to the organisation:

https://www.amd.com/en.html

Model release date:

March 2025

Availability

Base Model Data

Data sources laid out in model card.

https://huggingface.co/amd/Instella-3B-Long-Instruct

End User Model Data

Data sources laid out in model card.

https://huggingface.co/amd/Instella-3B-Long-Instruct

Base Model Weights

Weights published on HuggingFace.

https://huggingface.co/amd/Instella-3B

End User Model Weights

Weights published on HuggingFace.

https://huggingface.co/amd/Instella-3B-Long-Instruct

Training Code

Training code published on GitHub.

https://github.com/AMD-AIG-AIMA/Instella

Documentation

Code Documentation

Code is well-documented.

https://github.com/AMD-AIG-AIMA/Instella

Hardware Architecture

Hardware architecture outlined in preprint.

https://arxiv.org/pdf/2511.10628

Preprint

Preprint published on arXiv.

https://arxiv.org/pdf/2511.10628

Paper

No peer-reviewed paper found.

Modelcard

Model card contains the requisite detail.

https://huggingface.co/amd/Instella-3B-Long-Instruct

Datasheet

All datasets used are open and documented.

https://huggingface.co/amd/Instella-3B-Long-Instruct

Access

Licenses

Research-only RAIL-M5, not an OSI-approved license.

Is this information not up to date?

Contribute here ->

Supported by the Centre for Language Studies and the Dutch Research Council. Website design & development © 2024 by BSTN. This version of the index generated 09 April 2026, website content last updated 11 March 2026.