Hallucination

30 scenarios across 3 subcategories. 146 models scored.

Subcategories

Subcategory Avg Score Best Model Best Score
False Positives 75.58% o4 Mini High 98.59%
Content Invention 97.74% GPT-5.6 Sol (Reasoning) 100.00%
Output Corruption 99.65% GPT-5.6 Sol (Reasoning) 100.00%

Model Leaderboard

All models ranked by their Hallucination category score.

# Model Hallucination False Positives Content Invention Output Corruption Overall
1 o4 Mini High 99.53% 98.59% 100.00% 100.00% 88.78%
2 Z.AI GLM 5 Turbo 99.32% 97.97% 100.00% 99.99% 93.29%
3 Claude Opus 4.6 (Reasoning) 99.06% 97.19% 100.00% 100.00% 95.06%
4 Grok 4.5 (Reasoning, High) 99.06% 97.19% 100.00% 100.00% 94.12%
5 Grok 4.5 (Reasoning, Low) 99.01% 97.03% 100.00% 100.00% 90.94%
6 Z.AI GLM 5.2 (Reasoning, High) 98.99% 96.98% 100.00% 99.98% 93.41%
7 Z.AI GLM 5.1 98.84% 96.51% 100.00% 100.00% 93.74%
8 Gemma 4 26B (Reasoning) 98.79% 96.41% 100.00% 99.95% 89.02%
9 GPT-5.4 Mini (Reasoning, Low) 98.78% 96.36% 100.00% 99.98% 83.57%
10 Claude Opus 4.7 (Reasoning) 98.69% 96.09% 100.00% 99.98% 92.53%
11 GPT-5.6 Luna (Reasoning) 98.65% 95.94% 100.00% 100.00% 90.61%
12 o4 Mini 98.54% 95.63% 100.00% 100.00% 86.56%
13 Qwen 3.5 Plus (2026-04-20) 98.38% 96.25% 100.00% 98.89% 89.79%
14 GPT-5 Mini 98.28% 94.85% 100.00% 100.00% 91.31%
15 GPT-5.4 Mini (Reasoning) 98.11% 94.32% 100.00% 100.00% 89.82%
16 Z.AI GLM 5 98.04% 94.11% 100.00% 100.00% 89.60%
17 Grok 4.3 (Reasoning) 97.86% 93.59% 100.00% 100.00% 90.99%
18 MiniMax M3 97.83% 93.49% 100.00% 100.00% 90.45%
19 Gemini 2.5 Flash (Reasoning) 97.79% 93.44% 100.00% 99.94% 84.14%
20 GPT-OSS 120B 97.63% 92.97% 100.00% 99.92% 84.81%
21 Xiaomi MIMO v2.5 97.61% 92.92% 100.00% 99.93% 83.95%
22 ByteDance Seed 1.6 97.57% 92.71% 100.00% 100.00% 89.59%
23 GPT-5.2 97.55% 92.64% 100.00% 100.00% 89.45%
24 Qwen 3.6 27B 97.42% 92.33% 100.00% 99.95% 88.33%
25 Claude Opus 4.8 (Reasoning) 97.41% 92.24% 100.00% 100.00% 92.33%
26 Nemotron 3 Super 97.34% 92.03% 100.00% 100.00% 81.69%
27 Claude Opus 4.8 (Reasoning, Low) 97.33% 91.98% 100.00% 100.00% 91.89%
28 Qwen3.7 Max 97.15% 91.51% 100.00% 99.95% 94.55%
29 Xiaomi MIMO v2.5 Pro 97.02% 91.05% 100.00% 100.00% 86.05%
30 Gemini 2.5 Flash Lite (Reasoning) 96.99% 95.78% 95.24% 99.96% 83.10%
31 GPT-5.4 Nano (Reasoning) 96.98% 90.95% 100.00% 100.00% 80.02%
32 Claude Sonnet 4.6 (Reasoning) 96.98% 90.94% 100.00% 100.00% 93.64%
33 DeepSeek V4 Flash (Reasoning) 96.92% 94.32% 96.43% 100.00% 88.06%
34 Claude Opus 4.6 96.80% 90.40% 100.00% 100.00% 92.31%
35 Claude Sonnet 5 (Reasoning) 96.77% 90.31% 100.00% 100.00% 90.40%
36 Claude Sonnet 5 (Reasoning, Low) 96.44% 89.32% 100.00% 100.00% 90.16%
37 Gemma 4 31B (Reasoning) 96.37% 89.11% 100.00% 100.00% 89.64%
38 Grok 4.20 (Reasoning) 96.30% 88.89% 100.00% 100.00% 90.87%
39 GPT-5.1 96.22% 88.65% 100.00% 100.00% 90.73%
40 Qwen 3.6 Flash 96.13% 88.39% 100.00% 100.00% 89.31%
41 GPT-5 95.92% 87.76% 100.00% 100.00% 91.48%
42 GPT-5.6 Sol 95.87% 87.60% 100.00% 100.00% 90.76%
43 Qwen3.6 Max Preview 95.86% 87.58% 100.00% 100.00% 93.72%
44 GPT-5.4 (Reasoning, Low) 95.71% 87.14% 100.00% 100.00% 90.91%
45 GPT-5.4 Nano (Reasoning, Low) 95.71% 87.13% 100.00% 100.00% 77.46%
46 Gemini 3.5 Flash (Reasoning) 95.61% 86.82% 100.00% 100.00% 93.35%
47 GPT-4.1 95.60% 86.82% 100.00% 99.97% 86.82%
48 Aion 3.0 95.57% 86.82% 100.00% 99.88% 88.78%
49 DeepSeek V4 Pro (Reasoning) 95.25% 86.87% 100.00% 98.89% 89.28%
50 GPT-5.4 (Reasoning) 95.22% 85.65% 100.00% 100.00% 93.85%
51 Gemma 4 31B 95.09% 85.26% 100.00% 100.00% 85.23%
52 Claude Sonnet 4.6 94.99% 84.99% 100.00% 99.97% 90.66%
53 GPT-5.6 Terra (Reasoning) 94.98% 84.95% 100.00% 100.00% 92.49%
54 MiniMax M2.5 94.72% 84.66% 100.00% 99.50% 86.71%
55 MiniMax M2.7 94.69% 84.34% 100.00% 99.72% 86.23%
56 Gemini 3.1 Pro (Preview) 94.53% 83.59% 100.00% 100.00% 94.08%
57 MoonshotAI: Kimi K2.6 94.43% 83.28% 100.00% 100.00% 92.57%
58 Z.AI GLM 4.6 94.42% 84.38% 100.00% 98.89% 87.64%
59 Z.AI GLM 4.7 94.23% 82.73% 100.00% 99.96% 87.67%
60 Gemma 4 26B 94.22% 82.67% 100.00% 100.00% 84.89%
61 Aion 3.0 Mini 94.08% 88.15% 95.24% 98.85% 83.69%
62 Qwen 3.6 35B 94.07% 83.33% 100.00% 98.89% 87.66%
63 ByteDance Seed 2.0 Mini 94.07% 82.30% 100.00% 99.90% 85.69%
64 Ministral 3 8B 94.02% 82.06% 100.00% 100.00% 69.98%
65 MoonshotAI: Kimi K2.5 94.01% 82.03% 100.00% 100.00% 90.86%
66 Z.AI GLM 4.5 Air 94.00% 83.12% 100.00% 98.87% 80.74%
67 GPT-5.6 Luna 93.37% 80.10% 100.00% 100.00% 85.06%
68 Qwen 3.5 122B 93.29% 79.87% 100.00% 100.00% 90.32%
69 Qwen 3.5 27B 93.29% 79.86% 100.00% 100.00% 90.05%
70 Inception Mercury 2 93.22% 98.13% 81.55% 99.99% 81.99%
71 Mistral Small 4 (Reasoning) 93.18% 83.11% 96.43% 100.00% 79.48%
72 Aion 2.0 93.10% 86.41% 92.89% 100.00% 86.66%
73 Gemini 2.5 Pro 93.05% 79.16% 100.00% 100.00% 88.44%
74 GPT-5 Nano 92.99% 90.90% 88.10% 99.96% 80.16%
75 Claude Opus 4.7 92.95% 78.84% 100.00% 100.00% 89.90%
76 Gemini 3 Flash (Preview, Reasoning) 92.65% 77.99% 100.00% 99.97% 89.93%
77 GPT-5.5 (Reasoning) 92.12% 76.35% 100.00% 100.00% 93.72%
78 Claude Haiku 4.5 91.93% 75.80% 100.00% 100.00% 83.36%
79 Ministral 8B 91.89% 75.68% 100.00% 100.00% 63.77%
80 Z.AI GLM 4.5 91.83% 76.65% 100.00% 98.85% 84.95%
81 Qwen 3.5 9B 91.81% 81.32% 96.43% 97.67% 84.05%
82 GPT-5.6 Sol (Reasoning) 91.72% 75.16% 100.00% 100.00% 95.15%
83 GPT-5.5 (Reasoning, Low) 91.37% 74.11% 100.00% 100.00% 92.51%
84 Claude Opus 4.5 91.03% 73.09% 100.00% 100.00% 89.60%
85 ByteDance Seed 1.6 Flash 90.83% 83.03% 89.88% 99.57% 70.92%
86 GPT-5.6 Terra 90.63% 71.90% 100.00% 100.00% 88.23%
87 GPT-5.5 90.28% 70.85% 100.00% 100.00% 89.37%
88 GPT-4.1 Nano 90.07% 70.27% 100.00% 99.95% 69.90%
89 Claude Sonnet 4 90.04% 70.11% 100.00% 100.00% 87.64%
90 Qwen 3.5 397B A17B 90.04% 70.17% 100.00% 99.93% 91.09%
91 Z.AI GLM 4.7 Flash 90.00% 77.13% 92.86% 100.00% 82.21%
92 Gemini 3.5 Flash (Reasoning, Minimal) 89.55% 68.64% 100.00% 100.00% 85.88%
93 ByteDance Seed 2.0 Lite 89.50% 70.73% 100.00% 97.78% 84.27%
94 DeepSeek V4 Pro 89.36% 68.09% 100.00% 99.98% 82.05%
95 Qwen 3.5 35B 89.24% 74.87% 92.86% 100.00% 87.01%
96 Qwen 3 32B 89.06% 67.79% 100.00% 99.39% 79.37%
97 Mistral Medium 3.1 88.93% 66.78% 100.00% 100.00% 76.08%
98 Claude Sonnet 5 88.74% 66.23% 100.00% 100.00% 87.34%
99 Qwen 3.5 Flash 88.70% 73.21% 92.90% 99.98% 85.66%
100 Nemotron 3 Nano 88.55% 85.80% 82.14% 97.72% 74.50%
101 Arcee AI: Trinity Mini 88.35% 65.05% 100.00% 100.00% 67.68%
102 Gemini 2.5 Flash Lite 88.08% 64.28% 100.00% 99.96% 79.91%
103 Gemini 3.1 Flash Lite 87.86% 63.57% 100.00% 100.00% 85.09%
104 Claude Opus 4 87.84% 63.52% 100.00% 100.00% 87.22%
105 Grok 4.3 87.79% 63.41% 100.00% 99.96% 78.00%
106 Claude Sonnet 4.5 87.78% 63.35% 100.00% 100.00% 87.54%
107 DeepSeek V4 Flash 87.73% 63.20% 100.00% 100.00% 82.02%
108 Gemini 3.1 Flash Lite (Reasoning) 87.63% 62.89% 100.00% 100.00% 85.91%
109 Gemini 3.1 Flash Lite (Preview) 87.29% 61.89% 100.00% 99.97% 85.41%
110 Hermes 3 405B 87.02% 61.06% 100.00% 100.00% 80.80%
111 GPT-5.4 Nano 86.86% 60.58% 100.00% 100.00% 72.16%
112 Mistral Small 3.2 24B 86.84% 62.38% 100.00% 98.14% 77.36%
113 Qwen 3.5 Plus (2026-02-15) 86.62% 60.20% 100.00% 99.65% 86.17%
114 GPT-5.4 86.32% 58.96% 100.00% 100.00% 84.31%
115 DeepSeek V3.2 86.25% 58.75% 100.00% 100.00% 82.22%
116 GPT-4.1 Mini 86.23% 58.70% 100.00% 100.00% 81.40%
117 GPT-4o, Aug. 6th (temp=1) 86.00% 58.17% 100.00% 99.82% 81.28%
118 DeepSeek V3 (2024-12-26) 85.69% 57.07% 100.00% 100.00% 82.62%
119 Gemini 3 Flash (Preview) 85.61% 56.88% 100.00% 99.95% 85.47%
120 DeepSeek V3.1 85.55% 56.76% 100.00% 99.89% 82.35%
121 Qwen 2.5 72B 85.54% 58.36% 100.00% 98.28% 73.17%
122 Mistral Large 2 85.45% 56.35% 100.00% 100.00% 81.50%
123 GPT-4o Mini (temp=1) 85.28% 55.83% 100.00% 100.00% 77.82%
124 Gemini 2.5 Flash 85.27% 56.72% 100.00% 99.08% 80.61%
125 Cydonia 24B V4.1 85.22% 66.97% 91.07% 97.62% 72.68%
126 Mistral Large 3 85.18% 55.54% 100.00% 100.00% 84.29%
127 GPT-4o, Aug. 6th (temp=0) 83.33% 50.00% 100.00% 100.00% 82.18%
128 Ministral 3 14B 83.19% 49.57% 100.00% 100.00% 70.45%
129 DeepSeek-V2 Chat 82.48% 56.98% 90.48% 100.00% 84.09%
130 GPT-4o Mini (temp=0) 82.23% 46.70% 100.00% 100.00% 76.86%
131 GPT-5.4 Mini 81.85% 45.55% 100.00% 100.00% 80.45%
132 Mistral Small 4 81.83% 45.50% 100.00% 100.00% 75.23%
133 Gemma 3 27B 80.53% 41.59% 100.00% 100.00% 75.70%
134 Writer: Palmyra X5 80.49% 48.62% 92.86% 100.00% 78.11%
135 Grok 4.20 80.48% 41.45% 100.00% 99.99% 81.21%
136 Llama 3.1 70B 80.47% 61.33% 92.86% 87.24% 77.41%
137 WizardLM 2 8x22b 79.28% 57.02% 85.73% 95.09% 71.45%
138 Gemma 3 12B 77.13% 31.38% 100.00% 100.00% 76.07%
139 DeepSeek V3 (2025-03-24) 76.88% 47.98% 88.10% 94.58% 79.93%
140 Qwen3 235B A22B Instruct 2507 76.34% 43.34% 85.71% 99.97% 78.07%
141 Gemma 3 4B 75.94% 27.83% 100.00% 100.00% 66.33%
142 Ministral 3 3B 75.10% 36.00% 89.29% 100.00% 65.02%
143 Ministral 3B 73.79% 35.66% 85.71% 100.00% 59.25%
144 Hermes 3 70B 70.71% 53.99% 60.79% 97.34% 69.74%
145 Mistral NeMO 69.32% 31.81% 76.19% 99.97% 63.80%
146 Cohere Command R+ (Aug. 2024) 66.30% 61.00% 37.90% 100.00% 67.04%