Arcee AI: Trinity Large Thinking
Arcee AI: Trinity Large Thinking is arcee-ai's model. On the un-leaked meo benchmark it ranks #78 of 82 with 41.6% accuracy and an Effective Value (π) of 0.0000 β earning a correct answer for about $0.0108 in 102s each.
as of 2026-10-18
Per-domain accuracy
61 correct of 148 questions asked across 10 domains Β· answer rate 100% Β· macro accuracy 36.1%. Headline accuracy is micro-averaged per meoβs methodology; per-domain question counts (correct/asked) are shown so you can judge significance.
Efficiency
- Cost per correct
- $0.0108
- Seconds per correct
- 102
- Tokens per correct
- 11,608
- Correct answers
- 61 / 148
- Total benchmark cost
- $0.66
- Avg tokens / item
- 4,784
Aggregated third-party benchmarks
Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.
Model details
- Lab
- Arcee Ai
- Released
- 2026-04-01
- Weights
- Open
- Context
- 262K
- Max output
- 80K
- Knowledge cutoff
- β
- Price in
- $0.25/M
- Price out
- $0.80/M
- Modality
- text->text
Sources: openrouter Β· artificial_analysis Β· meo
FAQ
How accurate is Arcee AI: Trinity Large Thinking?
Arcee AI: Trinity Large Thinking scores 41.6% overall on the meo benchmark (rank #78 of 82), an un-leaked private holdout graded objective-first.
What is Arcee AI: Trinity Large Thinking's Effective Value?
Its Effective Value (π) is 0.0000 at N=10, Ο=1, Ξ΄=1.5 β a fused metric of accuracy, speed, cost, and error-cascade. π rankings shift with task depth.
How much does Arcee AI: Trinity Large Thinking cost per correct answer?
About $0.0108 per correct answer (11,608 tokens, 102 seconds each) at benchmark-time pricing via OpenRouter.
Is Arcee AI: Trinity Large Thinking open weights?
Yes β Arcee AI: Trinity Large Thinking is released with open weights.
Benchmark updates affecting this model
- 2026-06-09majorlogic_math_cs ground truth was broken β culled and rebuilt with solver-verified items
A 5-lab frontier jury revealed that 29 of 37 LLM-authored logic_math_cs items had broken ground truth: on most, five independent frontier models unanimously concluded the clue sets were contradictory while the stored answer claimed a number, and on eight, all five agreed on the SAME different number (author arithmetic errors). Models were being marked wrong for being right. The 29 were retired (full audit published in the repo) and the domain rebuilt with generator-as-oracle items: an embedded constraint solver PROVES each puzzle has exactly one solution (or provably none), verified by an independent exhaustive enumerator. A 3-round difficulty study found frontier models saturate consistent human-scale logic puzzles, so items use anonymous global constraints and carry a measured difficulty gradient.
- 2026-06-08majorCorrected base_bias grading (LLM-equivalence judge fix)
The LLM-equivalence grader uses a reasoning model whose verdict token budget was too small, so its YES/NO answer was emitted as empty and silently counted as a failure. This understated the base_bias domain to ~7% for every model when the true accuracy is ~85-90%. We raised the judge budget, excluded its reasoning tokens from the verdict, switched to verbosity-robust conclusion matching, and re-graded the whole board.
Embed this ranking
Show Arcee AI: Trinity Large Thinkingβs meo rank on your site β the badge links back here.
Related: all Arcee Ai models Β· Base-rate bias domain Β· best Effective Value Β· biggest movers Β· methodology Β· full leaderboard.