Neo

by Multimodal Art Projection

About the model:

Language model aiming to be completely open-source.

Model type:

Text

Model performance class:

Full

Link to the model:

https://huggingface.co/m-a-p/neo_7b_instruct_v0.1

Base models:

Neo-7B

End model:

Neo-7B-Instruct-v0.1

End model license:

Apache-2.0

About the organisation:

Open-source AI research community.

Link to the organisation:

https://m-a-p.ai/

Model release date:

May 2024

Availability

Base Model Data

https://huggingface.co/datasets/m-a-p/Matrix

End User Model Data

Some datasets listed in preprint but some missing; "we collect over 100k multiturn dialogue data sourced from real user conversations".

Base Model Weights

https://huggingface.co/m-a-p/neo_7b

End User Model Weights

https://huggingface.co/m-a-p/neo_7b_instruct_v0.1

Training Code

some code shared via github mostly based on NVIDIAs megatron

https://github.com/multimodal-art-projection/MAP-NEO

Documentation

Code Documentation

Hardware Architecture

Preprint

https://arxiv.org/abs/2405.19327

Paper

Modelcard

https://huggingface.co/m-a-p/neo_7b_instruct_v0.1

Datasheet

shared via preprint for base model

Access

Licenses

models under apache 2 code under MIT but not all datasets available

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.