Benchmarks · other
AA-PersonalityBench Warmth AA-PersonalityBench Warmth index: friendly, sincere, emotionally intelligent, with little sycophancy. Percentile rank (1 to 100) across the models AA evaluated, from pairwise LLM-judged trait Elos. Describes style, not correctness.
8 models · 8 results · data as of 2026-10-29 · artificialanalysis.ai ↗
1 Fable 5.1 97.5 2 GLM 5.3 90.1 3 Opus 5.5 82.7 4 GPT-6 Astra 72.8 5 GPT-6.1 Sol 62.9 6 MiMo V2.6 Pro 55.5 7 Grok 4.7 38.1 8 DeepSeek V4.1 Flash 1017 18.3 Fable 5.1 max
More from this benchmark 97 published measures Score vs any published measure 97 measures: sub-scores, splits, counts Trait Elo · Lexically varied Trait Elo · Eloquent Trait Elo · Literate Trait Elo · Natural sounding Trait Elo · Metaphorical Trait Elo · Structured Trait Elo · Scannable Trait Elo · Summarizes bottom line Trait Elo · Answer first Trait Elo · Information dense Trait Elo · Evidence based Trait Elo · Resource specific Trait Elo · Provides actionable guidance Trait Elo · Pragmatic Trait Elo · Truthful Trait Elo · Admits uncertainty Trait Elo · Disclaims with assumptions Trait Elo · Flags nuances Trait Elo · Sycophantic Trait Elo · Friendly Trait Elo · Humorous Trait Elo · Sincere Trait Elo · Emotionally intelligent Trait Elo · Direct Trait Elo · Confident Trait Elo · Opinionated Trait Elo · Constructively pushes back Trait Elo · Creative Trait Elo · Open ended Trait Elo · Slop Lexically varied · Lower Lexically varied · Upper Eloquent · Lower Eloquent · Upper Literate · Lower Literate · Upper Natural sounding · Lower Natural sounding · Upper Metaphorical · Lower Metaphorical · Upper Structured · Lower Structured · Upper Scannable · Lower Scannable · Upper Summarizes bottom line · Lower Summarizes bottom line · Upper Answer first · Lower Answer first · Upper Information dense · Lower Information dense · Upper Evidence based · Lower Evidence based · Upper Resource specific · Lower Resource specific · Upper Provides actionable guidance · Lower Provides actionable guidance · Upper Pragmatic · Lower Pragmatic · Upper Truthful · Lower Truthful · Upper Admits uncertainty · Lower Admits uncertainty · Upper Disclaims with assumptions · Lower Disclaims with assumptions · Upper Flags nuances · Lower Flags nuances · Upper Sycophantic · Lower Sycophantic · Upper Friendly · Lower Friendly · Upper Humorous · Lower Humorous · Upper Sincere · Lower Sincere · Upper Emotionally intelligent · Lower Emotionally intelligent · Upper Direct · Lower Direct · Upper Confident · Lower Confident · Upper Opinionated · Lower Opinionated · Upper Constructively pushes back · Lower Constructively pushes back · Upper Creative · Lower Creative · Upper Open ended · Lower Open ended · Upper Slop · Lower Slop · Upper Indexes · Expressiveness Indexes · Organization Indexes · Informativeness Indexes · Transparency Indexes · Conviction Indexes · Imagination Indexes · Slop ⌄
Breakdown darker is better within each column · click a column to sort BreakdownTrait Elo · 30 Indexes · 7 Trait intervals · Lexically varied · 2 Trait intervals · Eloquent · 2 Trait intervals · Literate · 2 Trait intervals · Natural sounding · 2 Trait intervals · Metaphorical · 2 Trait intervals · Structured · 2 Trait intervals · Scannable · 2 Trait intervals · Summarizes bottom line · 2 Trait intervals · Answer first · 2 Trait intervals · Information dense · 2 Trait intervals · Evidence based · 2 Trait intervals · Resource specific · 2 Trait intervals · Provides actionable guidance · 2 Trait intervals · Pragmatic · 2 Trait intervals · Truthful · 2 Trait intervals · Admits uncertainty · 2 Trait intervals · Disclaims with assumptions · 2 Trait intervals · Flags nuances · 2 Trait intervals · Sycophantic · 2 Trait intervals · Friendly · 2 Trait intervals · Humorous · 2 Trait intervals · Sincere · 2 Trait intervals · Emotionally intelligent · 2 Trait intervals · Direct · 2 Trait intervals · Confident · 2 Trait intervals · Opinionated · 2 Trait intervals · Constructively pushes back · 2 Trait intervals · Creative · 2 Trait intervals · Open ended · 2 Trait intervals · Slop · 2 Model Lexically varied Eloquent Literate Natural sounding Metaphorical Structured Scannable Summarizes bottom line Answer first Information dense Evidence based Resource specific Provides actionable guidance Pragmatic Truthful Admits uncertainty Disclaims with assumptions Flags nuances Sycophantic Friendly Humorous Sincere Emotionally intelligent Direct Confident Opinionated Constructively pushes back Creative Open ended Slop Fable 5.1 max, fallback 1,868 1,716 1,053 1,461 1,519 922.1 1,034 964.6 872.0 1,925 1,790 1,682 1,179 1,366 1,034 876.5 879.1 1,195 732.0 1,049 1,390 1,609 1,240 768.8 1,263 1,393 1,552 1,562 821.2 1,120 GLM 5.3 max 1,784 1,686 932.7 1,488 1,523 820.4 993.8 893.6 869.8 1,427 1,334 1,343 830.7 1,190 743.3 813.0 854.0 1,079 908.8 1,109 1,280 1,400 1,185 715.1 1,124 1,243 1,449 1,507 934.6 1,214 Opus 5.5 max, fallback 1,666 1,290 1,060 1,080 1,480 999.2 1,196 934.5 958.4 1,876 1,735 1,647 1,249 1,238 1,077 895.3 892.1 1,140 989.0 1,074 1,353 1,206 1,095 782.5 1,107 1,135 1,275 1,529 933.2 1,075 GPT-6 Astra max 1,042 1,060 993.4 1,041 1,022 1,017 960.1 955.8 941.5 991.9 1,040 1,000 1,050 1,062 1,039 1,009 975.9 1,023 985.0 1,004 1,085 1,085 1,052 936.9 947.5 990.1 1,010 1,036 1,039 1,006 GPT-6.1 Sol max 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 1,000 MiMo V2.6 Pro 856.2 1,077 643.4 693.1 1,325 712.4 1,545 944.2 652.4 1,042 917.3 1,394 784.1 647.3 456.3 796.7 723.2 792.6 862.2 1,038 1,078 657.3 948.9 672.6 808.7 777.3 1,224 1,012 1,045 1,629 Grok 4.7 xhigh 1,289 1,310 1,006 1,108 1,184 785.3 820.3 896.1 1,040 1,133 782.7 1,220 1,029 1,118 753.5 744.1 690.3 892.7 701.9 707.4 883.4 1,263 840.8 1,105 1,255 1,038 1,113 974.5 621.8 904.1 DeepSeek V4.1 Flash 1017 max 1,092 1,068 895.7 775.9 1,176 843.9 1,053 1,128 975.6 1,472 884.3 1,616 1,497 1,010 762.4 941.1 851.1 1,161 857.7 671.2 900.3 738.6 641.7 906.9 1,333 1,160 1,321 1,016 598.5 1,332