Skip to content

Benchmark

Arena ELO Text Factuality - Instruction Following

Instruction following benchmarks measure how closely models follow instructions. Higher scores are better.

Models measured
160
Scores recorded
179
Sources
1
IQ weight
2.15

Score distribution

1,436.3
Mean
1,439.3
Median
44.1
Std dev
1,377.3
10th percentile
1,486.3
90th percentile
1,280.8
Min
1,280.81,523.5
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.15 for chat models, measured over 160 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51GPT-6 LunaMax· LMArena data1,465.3
52Grok 4.5· LMArena data1,464.1
53GPT-6.1 TerraMax· LMArena data1,463.3
54Gemini 3.5 FlashMedium· LMArena data1,463.2
55GPT-6.2 SolMax· LMArena data1,462.2
56GPT-6.2 AstraMax· LMArena data1,461.8
57GPT-5.6 LunaxHigh· LMArena data1,461.6
58GPT 5.1High· LMArena data1,460.1
59ERNIE 5.1· LMArena data1,460
60Claude Sonnet 5High· LMArena data1,459.8
61Muse Spark 1.1· LMArena data1,458
62DeepSeek V4 ProHigh, 20260813· LMArena data1,456.8
63Kimi K2.6· LMArena data1,456.3
64GLM 5.1· LMArena data1,456.2
65DeepSeek V4 Pro· LMArena data1,455
66Step 5 Preview· LMArena data1,454.8
67Muse Spark· LMArena data1,453.9
68Grok 4.6High· LMArena data1,452.2
69MiMo V2 Pro· LMArena data1,450.9
70Qwen3.7 Plus· LMArena data1,449.8
71MiMo V2.6 Flash· LMArena data1,449.6
72Beam 1.0 501B A23B· LMArena data1,449.2
73Claude Opus 4.1Thinking 16K, 20250805· LMArena data1,449.2
74Qwen3.6 Max Preview· LMArena data1,448.1
75Claude Opus 4.120250805· LMArena data1,447.7
76DeepSeek-V4-Pro PreviewHigh· LMArena data1,447.5
77GPT 5.1· LMArena data1,446.1
78Gemini 3 Flash· LMArena data1,445.1
79Gemma 4 31B IT· LMArena data1,445
80Gemini 2.5 Pro· LMArena data1,445
81GLM 5· LMArena data1,444.8
82Kimi K2.5Instant· LMArena data1,444.2
83Inkling FP4· LMArena data1,443.5
84GPT-6 SolMax· LMArena data1,443.2
85MiMo V2.5· LMArena data1,443.1
86Grok 4.20Reasoning, 0309, beta· LMArena data1,441.8
87Qwen3.5 397B A17B· LMArena data1,441.1
88GPT-5.4 miniHigh· LMArena data1,440
89Gemma 4 26B A4B· LMArena data1,440
90Kimi K2.5Thinking· LMArena data1,439.3
91Hy3 (Deep Infra)· LMArena data1,439.2
92Grok 4.20 Multi-Agent0309, beta· LMArena data1,439.1
93Gemini 3.5 Flash Lite· LMArena data1,438.1
94GPT 5.2· LMArena data1,438
95MiniMax M3· LMArena data1,437.7
96GPT-5.2 Chat20260210· LMArena data1,437.2
97MiMo V2 Omni· LMArena data1,436.8
98Qwen 3.8 27B· LMArena data1,436.8
99GPT 5.2High· LMArena data1,434.6
100DeepSeek V4 Flash· LMArena data1,434.3

Showing 51–100 of 179 results.

Data credits: LMArena, CC BY 4.0 (licence)