LGCLFeb 19

Arcee Trinity Large Technical Report

arXiv:2602.17004v14 citationsh-index: 6Has Code
Originality Incremental advance
AI Analysis

This work addresses the problem of efficient large-scale model training for AI researchers and practitioners, presenting incremental improvements in architecture and training strategies.

The authors tackled the challenge of scaling sparse Mixture-of-Experts models by developing Trinity Large, Nano, and Mini, with Trinity Large achieving 400B total parameters and 13B activated per token, trained on 17 trillion tokens without loss spikes.

We present the technical report for Arcee Trinity Large, a sparse Mixture-of-Experts model with 400B total parameters and 13B activated per token. Additionally, we report on Trinity Nano and Trinity Mini, with Trinity Nano having 6B total parameters with 1B activated per token, Trinity Mini having 26B total parameters with 3B activated per token. The models' modern architecture includes interleaved local and global attention, gated attention, depth-scaled sandwich norm, and sigmoid routing for Mixture-of-Experts. For Trinity Large, we also introduce a new MoE load balancing strategy titled Soft-clamped Momentum Expert Bias Updates (SMEBU). We train the models using the Muon optimizer. All three models completed training with zero loss spikes. Trinity Nano and Trinity Mini were pre-trained on 10 trillion tokens, and Trinity Large was pre-trained on 17 trillion tokens. The model checkpoints are available at https://huggingface.co/arcee-ai.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes