Family of language models for text only.
SAPI-T
The SAPI-T (Sapiens-Text) is a family of multimodal language models with input-only multimodality. Models in this family accept inputs such as text, documents, images, audio, and video, but produce responses only in text format. They are models focused on lightness and inference speed.
SAPI-T4 [based on SAPI-4]
88 billion parameters with only 30 billion active.
Download
sapilm --get schatmodels/t4
Inference
sapilm --load sapi-t4
Contributing
We do not accept contributions that may result in changing the original code.
Make sure you are using the appropriate version.
License
This is proprietary software and its alteration and/or distribution without the developer's authorization is not permitted.