Arcee Trinity Large Breaks Cover

Arcee.ai unveils Trinity Large, a 400B-parameter Mixture-of-Experts model engineered for inference efficiency and enterprise long-context use, alongside smaller variants.

Feb 22 at 9:20 PM2 min read

Abstract representation of a neural network with nodes and connections, symbolizing the Arcee Trinity Large MoE architecture. — The intricate architecture of Arcee's Trinity Large, a 400B-parameter Mixture-of-Experts model.

Key Takeaways

1
Arcee's Trinity Large is a 400B-parameter open-weight MoE model designed for efficient enterprise deployment.
2
It features innovations like Soft-clamped Momentum Expert Bias Updates (SMEBU) and a custom multilingual tokenizer.
3
The model was pre-trained on 17 trillion tokens, including 8 trillion tokens of DatologyAI-curated synthetic data.

#Arcee.ai

#Trinity Large

#Mixture-of-Experts

#Large Language Models

#DatologyAI

Arcee Trinity Large Breaks Cover

AI Daily Digest

Arcee Trinity Large Breaks Cover

AI Daily Digest