Attention

Subcategory of Reasoning. 91 models scored.

Model Leaderboard

All models ranked by their Attention subcategory score.

# Model Attention Reasoning Overall
1 Claude Opus 4.5 98.41% 93.93% 89.69%
2 Claude Opus 4.6 (Reasoning) 98.10% 93.77% 95.02%
3 Claude Opus 4.6 97.78% 93.33% 92.35%
4 Grok 4 97.02% 96.01% 88.12%
5 Gemini 2.5 Pro 96.82% 96.91% 88.53%
6 Gemini 3 Flash (Preview, Reasoning) 96.60% 98.05% 90.50%
7 GPT-5 96.33% 95.67% 91.93%
8 Claude Sonnet 4.6 (Reasoning) 96.08% 92.76% 93.66%
9 Grok 4.1 Fast 96.05% 93.58% 89.55%
10 Gemini 3.1 Pro (Preview) 96.03% 96.01% 94.37%
11 Z.AI GLM 5 95.78% 95.89% 91.23%
12 Claude Sonnet 4.5 95.56% 92.50% 88.03%
13 Aion 2.0 95.49% 94.13% 89.21%
14 Gemini 3 Pro (Preview) 95.47% 95.24% 88.79%
15 GPT-5.1 95.29% 95.14% 92.54%
16 Qwen 3.5 397B A17B 95.13% 95.06% 91.73%
17 o4 Mini High 95.05% 95.02% 90.29%
18 Z.AI GLM 4.7 94.98% 94.99% 88.69%
19 Qwen 3.5 122B 94.87% 94.93% 91.53%
20 MoonshotAI: Kimi K2.5 94.83% 95.41% 91.04%
21 Grok 4 Fast 94.78% 94.89% 86.15%
22 Qwen 3.5 35B 94.75% 94.88% 88.00%
23 Gemini 3 Flash (Preview) 94.58% 94.79% 85.35%
24 Claude Sonnet 4 94.52% 94.48% 88.72%
25 Qwen 3.5 Flash 94.31% 94.66% 86.38%
26 GPT-5.2 93.58% 94.54% 90.26%
27 o4 Mini 93.40% 94.45% 88.35%
28 Claude Opus 4 93.24% 92.59% 87.69%
29 GPT-5 Mini 93.22% 94.36% 92.62%
30 Claude Sonnet 4.6 93.07% 88.48% 91.15%
31 ByteDance Seed 1.6 92.98% 91.49% 90.70%
32 Qwen 3.5 Plus (2026-02-15) 92.96% 93.45% 85.96%
33 Z.AI GLM 4.6 92.74% 95.12% 89.11%
34 Gemini 2.5 Flash (Reasoning) 91.73% 93.81% 86.51%
35 Gemini 2.5 Flash Lite (Reasoning) 91.22% 93.86% 85.75%
36 Claude 3.5 Sonnet 91.16% 90.30% 84.24%
37 Qwen 3.5 27B 90.46% 92.73% 90.85%
38 Claude 3.7 Sonnet 90.43% 89.94% 83.39%
39 Gemini 2.5 Flash 90.21% 92.60% 80.60%
40 Z.AI GLM 4.5 89.84% 91.03% 86.27%
41 Minimax M2.5 89.33% 92.42% 88.71%
42 Mistral Medium 3.1 89.19% 89.32% 77.83%
43 DeepSeek V3.2 88.87% 89.46% 82.25%
44 Mistral Large 2 88.62% 88.20% 82.41%
45 Mistral Large 3 88.45% 88.95% 85.43%
46 DeepSeek-V2 Chat 87.95% 88.70% 84.83%
47 GPT-4o, May 13th (temp=0) 87.71% 88.58% 85.36%
48 DeepSeek V3 (2025-03-24) 87.29% 88.45% 81.99%
49 DeepSeek V3 (2024-12-26) 87.02% 88.71% 83.68%
50 GPT-4.1 86.98% 88.46% 88.68%
51 Mistral Large 86.57% 76.31% 80.15%
52 Claude Haiku 4.5 86.08% 87.76% 85.14%
53 GPT-4o, Aug. 6th (temp=0) 85.73% 87.59% 82.45%
54 Stealth: Aurora Alpha 85.21% 90.11% 83.79%
55 GPT-4o, Aug. 6th (temp=1) 84.37% 86.91% 82.62%
56 GPT-5 Nano 84.22% 89.61% 82.60%
57 Z.AI GLM 4.7 Flash 83.99% 89.50% 84.82%
58 Writer: Palmyra X5 83.69% 86.57% 79.57%
59 Mistral Small Creative 83.04% 87.99% 73.27%
60 DeepSeek V3.1 82.96% 83.95% 82.39%
61 Hermes 3 405B 82.88% 85.58% 82.86%
62 GPT-4o, May 13th (temp=1) 82.52% 85.98% 83.80%
63 GPT-4.1 Mini 82.22% 85.83% 83.20%
64 ByteDance Seed 1.6 Flash 81.38% 86.52% 73.27%
65 Mistral Small 3.2 24B 79.53% 81.71% 78.60%
66 Gemma 3 27B 78.48% 86.74% 77.85%
67 Qwen 2.5 72B 77.42% 83.43% 75.46%
68 Llama 3.1 Nemotron 70B 77.15% 82.19% 74.70%
69 Ministral 3 14B 77.03% 83.24% 72.54%
70 Gemini 2.5 Flash Lite 76.60% 85.80% 81.08%
71 Ministral 3 8B 74.95% 71.64% 71.76%
72 Llama 3.1 70B 73.63% 79.31% 78.40%
73 Arcee AI: Trinity Mini 71.60% 76.94% 70.90%
74 Ministral 8B 71.56% 73.78% 64.87%
75 Arcee AI: Trinity Large (Preview) 70.59% 77.24% 73.33%
76 Claude 3.5 Haiku 70.51% 82.23% 83.73%
77 Hermes 3 70B 68.76% 79.08% 72.57%
78 GPT-4o Mini (temp=0) 68.07% 81.26% 78.29%
79 GPT-4o Mini (temp=1) 67.12% 80.28% 79.08%
80 Claude 3 Haiku 66.98% 77.94% 71.19%
81 Ministral 3B 66.56% 69.70% 61.29%
82 Cohere Command R+ (Aug. 2024) 66.37% 65.10% 69.03%
83 Ministral 3 3B 65.98% 71.88% 67.22%
84 WizardLM 2 8x22b 64.77% 67.36% 71.07%
85 Gemma 3 12B 63.84% 79.42% 78.41%
86 Gemma 3 4B 52.28% 73.64% 68.57%
87 Llama 3.1 8B 51.46% 69.12% 63.37%
88 GPT-4.1 Nano 46.04% 70.24% 71.94%
89 Rocinante 12B 33.73% 54.31% 54.55%
90 LFM2 24B 30.88% 54.88% 58.77%
91 Mistral NeMO 27.84% 57.59% 65.04%