Benchmarks

Model benchmarks

Every MohCitrus AI translation model is tested side by side on the same synthetic Discord chat messages and judged blind for meaning, natural wording, tone, formatting and speed. Every release gets its own report here.

Published reports. Select any model comparison to read the full results, category breakdown, example replies and methodology.

ComparisonDateMessagesOutcome
Terra vs Kiro20 November 202620 November 2026100Terra won 114–48 (38 ties)

Latest: Terra vs Kiro

Benchmark results. Terra and Kiro translated the same 100 Discord test messages across 24 target languages; across 200 blind head-to-head judgments, Terra won 114, Kiro won 48, and 38 were ties.

CategoryMeasureNotesTerraKiro
QualityQualityComparisons won†200 judged, 38 ties200 judged, 38 ties11448
Meaning kept†4.86 / 54.41 / 5
Natural wording†4.74 / 54.46 / 5
Tone kept†4.79 / 54.52 / 5
Formatting kept†4.97 / 54.88 / 5
ReliabilityReliabilityComplete replies199 of 200195 of 200
Mentions and links kept22 of 2220 of 22
Ordered-around messages resisted20 of 2018 of 20
SpeedSpeedTypical reply time‡Lower is betterLower is better592 ms748 ms
Slowest 1 in 20 replies‡Lower is betterLower is better1.48 s1.92 s

TerraKiro

Meaning kept

Terra4.86
Kiro4.41

Natural wording

Terra4.74
Kiro4.46

Tone kept

Terra4.79
Kiro4.52

Formatting kept

Terra4.97
Kiro4.88

05

Average blind-judge score out of 5 across all four quality criteria.

Try MohCitrus in your server

Free to start. Server Plus and Server Pro can switch between Terra and Kiro anytime in the dashboard.