zoom_in Click to enlarge

MAP-Neo

▲ ▼

description MAP-Neo Overview

MAP-Neo is a bilingual large language model developed by the OpenBMB team and researchers from the University of Science and Technology of China (USTC), released in 2024. It is designed as a fully open-source project, providing open access to its code, model weights, and training datasets. The model is pretrained on approximately 4.5 trillion tokens, with a strong emphasis on balancing English and Chinese language proficiency. It serves as a resource for the global AI research community, enabling transparent studies into model training and alignment.

help MAP-Neo FAQ

What is MAP-Neo?

MAP-Neo is a bilingual large language model series developed by the OpenBMB team and researchers from the University of Science and Technology of China. It was released as an open-source project in 2024.

How large is the main MAP-Neo model?

The main MAP-Neo research release has 7 billion parameters. Its training used about 4.5 trillion high-quality tokens, according to the model's technical paper.

Which languages does MAP-Neo target?

MAP-Neo is designed as a bilingual model, with Chinese and English as its primary languages. Its bilingual training focus is part of its distinction from English-only open models.

What does open source mean for MAP-Neo?

The MAP-Neo project provides public access to its code, model weights and training information. That transparency is intended to let researchers inspect how the model was built rather than treating it as a closed commercial service.

swap_horiz

Looking for MAP-Neo alternatives? Compare top competitors ranked & scored

arrow_forward