False Positives

Subcategory of Hallucination. 154 models scored.

Model Leaderboard

All models ranked by their False Positives subcategory score.

# Model False Positives Hallucination Overall
1 Nemotron 3 Super 99.06% 99.69% 84.56%
2 Inception Mercury 98.13% 95.08% 79.50%
3 o4 Mini High 97.19% 99.06% 90.29%
4 Grok 4.1 Fast 97.08% 99.02% 89.55%
5 GPT-5.4 Nano (Reasoning, Low) 97.08% 99.03% 79.48%
6 o4 Mini 96.25% 98.75% 88.35%
7 Inception Mercury 2 96.25% 92.60% 83.85%
8 Z.AI GLM 5 Turbo 95.94% 98.64% 94.27%
9 GPT-5.1 95.31% 98.44% 92.54%
10 GPT-5.4 Mini (Reasoning, Low) 95.31% 98.43% 85.75%
11 Claude Opus 4.6 (Reasoning) 94.38% 98.13% 95.02%
12 ByteDance Seed 1.6 Flash 94.11% 94.52% 73.27%
13 GPT-5.4 Nano (Reasoning) 93.85% 97.95% 81.36%
14 Z.AI GLM 5 93.23% 97.74% 91.23%
15 GPT-5 Mini 93.12% 97.71% 92.62%
16 Z.AI GLM 5.1 93.02% 97.67% 94.37%
17 Gemma 4 26B (Reasoning) 92.81% 97.59% 91.49%
18 Qwen 3.5 Plus (2026-04-20) 92.50% 97.13% 91.51%
19 GPT-5 Nano 92.50% 93.52% 82.60%
20 Grok 4.3 (Reasoning) 92.19% 97.40% 93.60%
21 Claude Opus 4.7 (Reasoning) 92.19% 97.39% 93.23%
22 Gemini 2.5 Flash Lite (Reasoning) 91.56% 95.59% 85.75%
23 MiniMax M2.7 89.69% 96.47% 89.10%
24 Grok 4.20 (Beta, Reasoning) 88.85% 96.28% 91.49%
25 GPT-5.4 Mini (Reasoning) 88.65% 96.22% 90.65%
26 DeepSeek V4 Flash (Reasoning) 88.65% 95.02% 89.01%
27 Aion 2.0 87.81% 93.57% 89.21%
28 Nemotron 3 Nano 87.60% 89.16% 77.73%
29 Qwen 3.6 27B 87.24% 95.73% 89.72%
30 MiniMax M3 86.98% 95.66% 90.88%
31 Grok 4.20 (Reasoning) 86.98% 95.66% 91.39%
32 Gemini 2.5 Flash (Reasoning) 86.87% 95.60% 86.51%
33 GPT-OSS 120B 85.94% 95.29% 86.44%
34 Xiaomi MIMO v2.5 85.83% 95.25% 85.05%
35 GPT-4.1 85.76% 95.24% 88.68%
36 Xiaomi MIMO v2.5 Pro 85.52% 95.17% 87.36%
37 ByteDance Seed 1.6 85.42% 95.14% 90.70%
38 GPT-5.2 85.28% 95.09% 90.26%
39 Claude Opus 4.8 (Reasoning) 84.48% 94.83% 92.22%
40 Stealth: Healer Alpha 84.06% 94.67% 85.93%
41 Claude Opus 4.8 (Reasoning, Low) 83.96% 94.65% 92.14%
42 Gemma 4 31B (Reasoning) 83.23% 94.41% 91.71%
43 Qwen3.7 Max 83.02% 94.32% 95.75%
44 Mistral Small 4 (Reasoning) 82.50% 92.98% 82.39%
45 Claude Sonnet 4.6 (Reasoning) 81.88% 93.96% 93.66%
46 Claude Opus 4.6 80.81% 93.60% 92.35%
47 Z.AI GLM 4.5 Air 79.34% 92.74% 83.12%
48 MiniMax M2.5 79.32% 92.94% 88.71%
49 Ministral 3 8B 77.55% 92.52% 71.76%
50 GPT-5.4 (Reasoning, Low) 76.88% 92.29% 91.41%
51 Qwen 3.6 Flash 76.77% 92.26% 90.65%
52 Z.AI GLM 4.7 Flash 76.72% 89.86% 84.82%
53 GPT-5 75.52% 91.84% 91.93%
54 Qwen3.6 Max Preview 75.16% 91.72% 94.54%
55 ByteDance Seed 2.0 Mini 73.98% 91.29% 86.91%
56 DeepSeek V4 Pro (Reasoning) 73.75% 90.88% 90.10%
57 Gemini 3.5 Flash (Reasoning) 73.65% 91.22% 94.08%
58 Arcee AI: Trinity Mini 73.37% 91.12% 70.90%
59 Grok 4 Fast 73.28% 91.09% 86.15%
60 Stealth: Hunter Alpha 72.33% 90.78% 87.34%
61 Qwen 3.6 35B 71.67% 90.19% 89.05%
62 LFM2 24B 71.58% 90.53% 58.77%
63 Z.AI GLM 4.6 71.35% 90.08% 89.11%
64 GPT-5.4 (Reasoning) 71.30% 90.43% 93.24%
65 Gemma 4 31B 70.52% 90.17% 86.91%
66 Claude Sonnet 4.6 69.98% 89.98% 91.15%
67 Qwen 3 32B 69.27% 89.56% 82.21%
68 Grok 4 68.35% 89.45% 88.12%
69 GPT-5.4 Nano 67.86% 89.29% 74.40%
70 Ministral 8B 67.57% 89.19% 64.87%
71 Gemini 3.1 Pro (Preview) 67.19% 89.06% 94.37%
72 MoonshotAI: Kimi K2.6 66.56% 88.85% 92.31%
73 Z.AI GLM 4.7 65.47% 88.47% 88.69%
74 Gemma 4 26B 65.34% 88.45% 85.84%
75 Gemini 3 Pro (Preview) 64.69% 88.23% 88.79%
76 MoonshotAI: Kimi K2.5 64.06% 88.02% 91.04%
77 GPT-4.1 Nano 63.25% 87.73% 71.94%
78 Qwen 3.5 9B 62.64% 85.58% 86.05%
79 Z.AI GLM 4.5 62.31% 87.05% 86.27%
80 Qwen 3.5 122B 59.74% 86.58% 91.53%
81 Qwen 3.5 27B 59.73% 86.58% 90.85%
82 Gemini 2.5 Pro 58.32% 86.11% 88.53%
83 Claude Opus 4.7 57.68% 85.89% 89.93%
84 Gemini 3 Flash (Preview, Reasoning) 55.98% 85.32% 90.50%
85 Cydonia 24B V4.1 55.90% 81.53% 75.09%
86 Cohere Command R+ (Aug. 2024) 55.29% 64.40% 69.03%
87 GPT-5.5 (Reasoning, Low) 53.23% 84.41% 92.59%
88 GPT-5.5 (Reasoning) 52.71% 84.24% 92.98%
89 Claude Haiku 4.5 51.59% 83.86% 85.14%
90 Qwen 3.5 35B 49.74% 80.87% 88.00%
91 Qwen 3.5 Flash 49.01% 80.63% 86.38%
92 Qwen 3.5 397B A17B 46.35% 82.10% 91.73%
93 Mistral Medium 3.1 46.27% 82.09% 77.83%
94 Claude Opus 4.5 46.19% 82.06% 89.69%
95 Skyfall 36B V2 44.87% 66.29% 65.76%
96 GPT-4.1 Mini 43.42% 81.14% 83.20%
97 Hermes 3 70B 42.95% 67.03% 72.57%
98 GPT-5.5 41.70% 80.57% 89.09%
99 ByteDance Seed 2.0 Lite 41.46% 79.75% 84.80%
100 Qwen 2.5 72B 40.41% 79.56% 75.46%
101 Claude Sonnet 4 40.23% 80.08% 88.72%
102 Ministral 3 14B 39.97% 79.99% 72.54%
103 Hermes 3 405B 39.09% 79.70% 82.86%
104 GPT-4o, Aug. 6th (temp=1) 38.76% 79.53% 82.62%
105 Gemini 3.5 Flash (Reasoning, Minimal) 37.28% 79.09% 86.47%
106 DeepSeek V4 Pro 36.18% 78.72% 82.63%
107 GPT-5.4 Mini 35.21% 78.40% 82.43%
108 Grok 4.20 (Beta) 34.84% 78.28% 83.85%
109 Mistral Large 3 34.52% 78.17% 85.43%
110 Mistral Large 2 33.60% 77.87% 82.41%
111 Mistral Large 32.50% 77.50% 80.15%
112 Grok 4.3 31.51% 77.16% 78.66%
113 GPT-4o Mini (temp=1) 30.35% 76.78% 79.08%
114 WizardLM 2 8x22b 29.74% 70.19% 71.06%
115 Arcee AI: Trinity Large (Preview) 29.41% 76.47% 73.33%
116 Llama 3.1 70B 29.26% 69.78% 78.40%
117 Claude 3.5 Sonnet 28.94% 76.31% 84.24%
118 Llama 3.1 Nemotron 70B 28.79% 74.99% 74.70%
119 Mistral Small 3.2 24B 28.78% 75.64% 78.58%
120 Gemini 2.5 Flash Lite 28.56% 76.17% 81.08%
121 Rocinante 12B 27.40% 52.07% 54.54%
122 Gemini 3.1 Flash Lite 27.15% 75.72% 85.75%
123 Claude Opus 4 27.04% 75.68% 87.69%
124 Claude Sonnet 4.5 26.70% 75.57% 88.03%
125 Ministral 3B 26.52% 70.75% 61.29%
126 DeepSeek V4 Flash 26.40% 75.47% 82.02%
127 Gemini 3.1 Flash Lite (Reasoning) 25.79% 75.26% 86.41%
128 Claude 3.7 Sonnet 25.54% 75.18% 83.39%
129 Gemini 3.1 Flash Lite (Preview) 23.78% 74.58% 85.87%
130 Qwen3 235B A22B Instruct 2507 23.76% 69.82% 80.10%
131 Mistral Small Creative 23.38% 74.46% 73.27%
132 Writer: Palmyra X5 23.27% 72.04% 79.57%
133 Ministral 3 3B 22.08% 70.45% 67.22%
134 GPT-5.4 21.34% 73.78% 84.32%
135 GPT-4o Mini (temp=0) 20.67% 73.56% 78.29%
136 Qwen 3.5 Plus (2026-02-15) 20.40% 73.35% 85.96%
137 Mistral Small 4 20.22% 73.41% 76.46%
138 GPT-4o, May 13th (temp=1) 20.21% 73.40% 83.80%
139 GPT-4o, Aug. 6th (temp=0) 20.06% 73.35% 82.45%
140 Claude 3 Haiku 19.94% 60.81% 71.19%
141 DeepSeek V3 (2024-12-26) 19.33% 73.11% 83.68%
142 Grok 4.20 19.05% 73.01% 81.70%
143 DeepSeek V3 (2025-03-24) 18.55% 67.07% 81.99%
144 DeepSeek V3.1 18.52% 72.80% 82.39%
145 Llama 3.1 8B 18.43% 43.27% 63.35%
146 DeepSeek-V2 Chat 17.97% 69.48% 84.83%
147 DeepSeek V3.2 17.51% 72.50% 82.25%
148 Gemini 2.5 Flash 16.03% 71.70% 80.60%
149 Gemini 3 Flash (Preview) 13.77% 71.24% 85.35%
150 Mistral NeMO 11.73% 62.63% 65.04%
151 GPT-4o, May 13th (temp=0) 9.29% 69.76% 85.36%
152 Gemma 3 12B 7.45% 69.15% 78.41%
153 Gemma 3 27B 6.23% 68.74% 77.85%
154 Gemma 3 4B 2.80% 67.60% 68.57%