Inference
EN

Trinity-Large-Preview vs Trinity-Large-Thinking

Change models
Trinity-Large-PreviewArcee AI
Trinity-Large-ThinkingArcee AI
IndexAll categories, out of 100.
42.2 ±7.6
43.5 ±7.2 best
GapIndex points below the top.
−1.3 · no clear gap
Top
CoverageIndex weight that has results.
75%
75%
SpeedOutput tokens per second.
—
138/s
EloLMArena rating and rank.
1339
1340 best
ReleasedMonth of release.
2026-01
2026-03
Input / 1MUS dollars per 1M input tokens.
$0.25
$0.25
Output / 1MUS dollars per 1M output tokens.
$1.00
$0.80 best
1K chats2K tokens in, 500 out, each.
$1.00
$0.90 best
1K documents20K tokens in, 1K out, each.
$6.00
$5.80 best
1K agent tasks200K tokens in, 20K out, each.
$70.00
$66.00 best
ContextMaximum tokens in one request.
512K best
262K
AgenticMulti-step tasks with tools.
48.3 best
44.0
CodingCode writing and repair.
48.8 best
48.3
ReasoningLogic problems and puzzles.
36.0
36.0
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
39.0
45.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
41.6 best
39.8
MathMath problems.
—
—

Benchmarks

11 shared
Only Trinity-Large-Preview: 1Only Trinity-Large-Thinking: 7
Trinity-Large-Preview
Trinity-Large-Thinking
Artificial Analysis Humanity's Last ExamKnowledge
15.8%123 / 222
15.8%123 / 222
Artificial Analysis Long Context ReasoningReasoning
38.0%169 / 217
38.0%169 / 217
Critical Physics TasksReasoning
0.9%124 / 214
0.9%124 / 214
Artificial Analysis GPQA DiamondKnowledge
75.2%116 / 199
75.2%116 / 199
τ²-Bench Tool-Agent-User EvaluationAgentic
90.1%40 / 149
90.1%40 / 149
Artificial Analysis IFBenchInstruction
56.3%68 / 146
56.3%68 / 146
GDPval-AA normalizedAgentic
0.0%101 / 133
0.0%101 / 133
Artificial Analysis SciCodeCoding
40.6%84 / 114
40.6%84 / 114
GPQA Diamond · BenchLMKnowledge
63.3%77 / 84
76.3% best67 / 84
MMLU-Pro first-party comparison snapshotKnowledge
75.2%6 / 6
83.4% best4 / 6
AIME25 first-party comparison snapshotMath
24.0%
96.3%