trinity-large-preview: benchmark results

Arcee AI's Trinity Large preview, a ~400B model (January 2026). Provider: Arcee AI. Released 2026-01-27. Access: Open.

Unified ELO 1526 ± 1, rank #564 of 1392 rated models, from 100 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AGC-Bench - fann_or_flop2.3Dataset z-score98.8
AGC-Bench - science_analogies1.94Dataset z-score93.8
UGI Leaderboard52.11UGI Score93.3
AGC-Bench - ocw_connections1.08Dataset z-score90.9
AGC-Bench - futuregen1.2Dataset z-score90
UGI - Natural Intelligence41.34NatInt Score87.2
AGC-Bench - story_generation_rocstories0.89Dataset z-score86.6
CringeBench0.76Cringe Score (0-10, lower is better)80.3
AGC-Bench - hypobench0.35Dataset z-score79.3
AGC-Bench - pun_eval0.67Dataset z-score79.3
UGI - Writing40.72Writing Score75.5
Vectara Hallucination Leaderboard93.1Factual Consistency Rate (%)72.6

Interactive version: theaggregate.ai/model?slug=trinity-large-preview · How It Works · Data refreshed daily, snapshot 2026-09-05.