Output Corruption
Subcategory of Hallucination. 146 models scored.
Model Leaderboard
All models ranked by their Output Corruption subcategory score.
| # | Model | Output Corruption | Hallucination | Overall |
|---|---|---|---|---|
| 1 | GPT-5.6 Sol (Reasoning) | 100.00% | 91.72% | 95.15% |
| 2 | Claude Opus 4.6 (Reasoning) | 100.00% | 99.06% | 95.06% |
| 3 | Grok 4.5 (Reasoning, High) | 100.00% | 99.06% | 94.12% |
| 4 | Gemini 3.1 Pro (Preview) | 100.00% | 94.53% | 94.08% |
| 5 | GPT-5.4 (Reasoning) | 100.00% | 95.22% | 93.85% |
| 6 | Z.AI GLM 5.1 | 100.00% | 98.84% | 93.74% |
| 7 | GPT-5.5 (Reasoning) | 100.00% | 92.12% | 93.72% |
| 8 | Claude Sonnet 4.6 (Reasoning) | 100.00% | 96.98% | 93.64% |
| 9 | Gemini 3.5 Flash (Reasoning) | 100.00% | 95.61% | 93.35% |
| 10 | MoonshotAI: Kimi K2.6 | 100.00% | 94.43% | 92.57% |
| 11 | GPT-5.5 (Reasoning, Low) | 100.00% | 91.37% | 92.51% |
| 12 | GPT-5.6 Terra (Reasoning) | 100.00% | 94.98% | 92.49% |
| 13 | Claude Opus 4.8 (Reasoning) | 100.00% | 97.41% | 92.33% |
| 14 | Claude Opus 4.6 | 100.00% | 96.80% | 92.31% |
| 15 | Claude Opus 4.8 (Reasoning, Low) | 100.00% | 97.33% | 91.89% |
| 16 | GPT-5 | 100.00% | 95.92% | 91.48% |
| 17 | GPT-5 Mini | 100.00% | 98.28% | 91.31% |
| 18 | Grok 4.3 (Reasoning) | 100.00% | 97.86% | 90.99% |
| 19 | Grok 4.5 (Reasoning, Low) | 100.00% | 99.01% | 90.94% |
| 20 | GPT-5.4 (Reasoning, Low) | 100.00% | 95.71% | 90.91% |
| 21 | Grok 4.20 (Reasoning) | 100.00% | 96.30% | 90.87% |
| 22 | MoonshotAI: Kimi K2.5 | 100.00% | 94.01% | 90.86% |
| 23 | GPT-5.6 Sol | 100.00% | 95.87% | 90.76% |
| 24 | GPT-5.1 | 100.00% | 96.22% | 90.73% |
| 25 | GPT-5.6 Luna (Reasoning) | 100.00% | 98.65% | 90.61% |
| 26 | MiniMax M3 | 100.00% | 97.83% | 90.45% |
| 27 | Claude Sonnet 5 (Reasoning) | 100.00% | 96.77% | 90.40% |
| 28 | Qwen 3.5 122B | 100.00% | 93.29% | 90.32% |
| 29 | Claude Sonnet 5 (Reasoning, Low) | 100.00% | 96.44% | 90.16% |
| 30 | Qwen 3.5 27B | 100.00% | 93.29% | 90.05% |
| 31 | Claude Opus 4.7 | 100.00% | 92.95% | 89.90% |
| 32 | Gemma 4 31B (Reasoning) | 100.00% | 96.37% | 89.64% |
| 33 | Claude Opus 4.5 | 100.00% | 91.03% | 89.60% |
| 34 | ByteDance Seed 1.6 | 100.00% | 97.57% | 89.59% |
| 35 | GPT-5.2 | 100.00% | 97.55% | 89.45% |
| 36 | GPT-5.5 | 100.00% | 90.28% | 89.37% |
| 37 | Qwen 3.6 Flash | 100.00% | 96.13% | 89.31% |
| 38 | o4 Mini High | 100.00% | 99.53% | 88.78% |
| 39 | Gemini 2.5 Pro | 100.00% | 93.05% | 88.44% |
| 40 | GPT-5.6 Terra | 100.00% | 90.63% | 88.23% |
| 41 | DeepSeek V4 Flash (Reasoning) | 100.00% | 96.92% | 88.06% |
| 42 | Claude Sonnet 4 | 100.00% | 90.04% | 87.64% |
| 43 | Claude Sonnet 4.5 | 100.00% | 87.78% | 87.54% |
| 44 | Claude Sonnet 5 | 100.00% | 88.74% | 87.34% |
| 45 | Claude Opus 4 | 100.00% | 87.84% | 87.22% |
| 46 | Qwen 3.5 35B | 100.00% | 89.24% | 87.01% |
| 47 | Aion 2.0 | 100.00% | 93.10% | 86.66% |
| 48 | o4 Mini | 100.00% | 98.54% | 86.56% |
| 49 | Xiaomi MIMO v2.5 Pro | 100.00% | 97.02% | 86.05% |
| 50 | Gemini 3.1 Flash Lite (Reasoning) | 100.00% | 87.63% | 85.91% |
| 51 | Gemini 3.5 Flash (Reasoning, Minimal) | 100.00% | 89.55% | 85.88% |
| 52 | Gemma 4 31B | 100.00% | 95.09% | 85.23% |
| 53 | Gemini 3.1 Flash Lite | 100.00% | 87.86% | 85.09% |
| 54 | GPT-5.6 Luna | 100.00% | 93.37% | 85.06% |
| 55 | Gemma 4 26B | 100.00% | 94.22% | 84.89% |
| 56 | GPT-5.4 | 100.00% | 86.32% | 84.31% |
| 57 | Mistral Large 3 | 100.00% | 85.18% | 84.29% |
| 58 | DeepSeek-V2 Chat | 100.00% | 82.48% | 84.09% |
| 59 | Claude Haiku 4.5 | 100.00% | 91.93% | 83.36% |
| 60 | DeepSeek V3 (2024-12-26) | 100.00% | 85.69% | 82.62% |
| 61 | DeepSeek V3.2 | 100.00% | 86.25% | 82.22% |
| 62 | Z.AI GLM 4.7 Flash | 100.00% | 90.00% | 82.21% |
| 63 | GPT-4o, Aug. 6th (temp=0) | 100.00% | 83.33% | 82.18% |
| 64 | DeepSeek V4 Flash | 100.00% | 87.73% | 82.02% |
| 65 | Nemotron 3 Super | 100.00% | 97.34% | 81.69% |
| 66 | Mistral Large 2 | 100.00% | 85.45% | 81.50% |
| 67 | GPT-4.1 Mini | 100.00% | 86.23% | 81.40% |
| 68 | Hermes 3 405B | 100.00% | 87.02% | 80.80% |
| 69 | GPT-5.4 Mini | 100.00% | 81.85% | 80.45% |
| 70 | GPT-5.4 Nano (Reasoning) | 100.00% | 96.98% | 80.02% |
| 71 | Mistral Small 4 (Reasoning) | 100.00% | 93.18% | 79.48% |
| 72 | Writer: Palmyra X5 | 100.00% | 80.49% | 78.11% |
| 73 | GPT-4o Mini (temp=1) | 100.00% | 85.28% | 77.82% |
| 74 | GPT-5.4 Nano (Reasoning, Low) | 100.00% | 95.71% | 77.46% |
| 75 | GPT-4o Mini (temp=0) | 100.00% | 82.23% | 76.86% |
| 76 | Mistral Medium 3.1 | 100.00% | 88.93% | 76.08% |
| 77 | Gemma 3 12B | 100.00% | 77.13% | 76.07% |
| 78 | Gemma 3 27B | 100.00% | 80.53% | 75.70% |
| 79 | Mistral Small 4 | 100.00% | 81.83% | 75.23% |
| 80 | GPT-5.4 Nano | 100.00% | 86.86% | 72.16% |
| 81 | Ministral 3 14B | 100.00% | 83.19% | 70.45% |
| 82 | Ministral 3 8B | 100.00% | 94.02% | 69.98% |
| 83 | Arcee AI: Trinity Mini | 100.00% | 88.35% | 67.68% |
| 84 | Cohere Command R+ (Aug. 2024) | 100.00% | 66.30% | 67.04% |
| 85 | Gemma 3 4B | 100.00% | 75.94% | 66.33% |
| 86 | Ministral 3 3B | 100.00% | 75.10% | 65.02% |
| 87 | Ministral 8B | 100.00% | 91.89% | 63.77% |
| 88 | Ministral 3B | 100.00% | 73.79% | 59.25% |
| 89 | GPT-5.4 Mini (Reasoning) | 100.00% | 98.11% | 89.82% |
| 90 | Qwen3.6 Max Preview | 100.00% | 95.86% | 93.72% |
| 91 | Z.AI GLM 5 | 100.00% | 98.04% | 89.60% |
| 92 | Z.AI GLM 5 Turbo | 99.99% | 99.32% | 93.29% |
| 93 | Grok 4.20 | 99.99% | 80.48% | 81.21% |
| 94 | Inception Mercury 2 | 99.99% | 93.22% | 81.99% |
| 95 | Z.AI GLM 5.2 (Reasoning, High) | 99.98% | 98.99% | 93.41% |
| 96 | Claude Opus 4.7 (Reasoning) | 99.98% | 98.69% | 92.53% |
| 97 | Qwen 3.5 Flash | 99.98% | 88.70% | 85.66% |
| 98 | DeepSeek V4 Pro | 99.98% | 89.36% | 82.05% |
| 99 | GPT-5.4 Mini (Reasoning, Low) | 99.98% | 98.78% | 83.57% |
| 100 | GPT-4.1 | 99.97% | 95.60% | 86.82% |
| 101 | Gemini 3 Flash (Preview, Reasoning) | 99.97% | 92.65% | 89.93% |
| 102 | Qwen3 235B A22B Instruct 2507 | 99.97% | 76.34% | 78.07% |
| 103 | Claude Sonnet 4.6 | 99.97% | 94.99% | 90.66% |
| 104 | Mistral NeMO | 99.97% | 69.32% | 63.80% |
| 105 | Gemini 3.1 Flash Lite (Preview) | 99.97% | 87.29% | 85.41% |
| 106 | Gemini 2.5 Flash Lite (Reasoning) | 99.96% | 96.99% | 83.10% |
| 107 | Grok 4.3 | 99.96% | 87.79% | 78.00% |
| 108 | GPT-5 Nano | 99.96% | 92.99% | 80.16% |
| 109 | Z.AI GLM 4.7 | 99.96% | 94.23% | 87.67% |
| 110 | Gemini 2.5 Flash Lite | 99.96% | 88.08% | 79.91% |
| 111 | Gemini 3 Flash (Preview) | 99.95% | 85.61% | 85.47% |
| 112 | Qwen3.7 Max | 99.95% | 97.15% | 94.55% |
| 113 | Gemma 4 26B (Reasoning) | 99.95% | 98.79% | 89.02% |
| 114 | Qwen 3.6 27B | 99.95% | 97.42% | 88.33% |
| 115 | GPT-4.1 Nano | 99.95% | 90.07% | 69.90% |
| 116 | Gemini 2.5 Flash (Reasoning) | 99.94% | 97.79% | 84.14% |
| 117 | Qwen 3.5 397B A17B | 99.93% | 90.04% | 91.09% |
| 118 | Xiaomi MIMO v2.5 | 99.93% | 97.61% | 83.95% |
| 119 | GPT-OSS 120B | 99.92% | 97.63% | 84.81% |
| 120 | ByteDance Seed 2.0 Mini | 99.90% | 94.07% | 85.69% |
| 121 | DeepSeek V3.1 | 99.89% | 85.55% | 82.35% |
| 122 | Aion 3.0 | 99.88% | 95.57% | 88.78% |
| 123 | GPT-4o, Aug. 6th (temp=1) | 99.82% | 86.00% | 81.28% |
| 124 | MiniMax M2.7 | 99.72% | 94.69% | 86.23% |
| 125 | Qwen 3.5 Plus (2026-02-15) | 99.65% | 86.62% | 86.17% |
| 126 | ByteDance Seed 1.6 Flash | 99.57% | 90.83% | 70.92% |
| 127 | MiniMax M2.5 | 99.50% | 94.72% | 86.71% |
| 128 | Qwen 3 32B | 99.39% | 89.06% | 79.37% |
| 129 | Gemini 2.5 Flash | 99.08% | 85.27% | 80.61% |
| 130 | Qwen 3.5 Plus (2026-04-20) | 98.89% | 98.38% | 89.79% |
| 131 | DeepSeek V4 Pro (Reasoning) | 98.89% | 95.25% | 89.28% |
| 132 | Qwen 3.6 35B | 98.89% | 94.07% | 87.66% |
| 133 | Z.AI GLM 4.6 | 98.89% | 94.42% | 87.64% |
| 134 | Z.AI GLM 4.5 Air | 98.87% | 94.00% | 80.74% |
| 135 | Aion 3.0 Mini | 98.85% | 94.08% | 83.69% |
| 136 | Z.AI GLM 4.5 | 98.85% | 91.83% | 84.95% |
| 137 | Qwen 2.5 72B | 98.28% | 85.54% | 73.17% |
| 138 | Mistral Small 3.2 24B | 98.14% | 86.84% | 77.36% |
| 139 | ByteDance Seed 2.0 Lite | 97.78% | 89.50% | 84.27% |
| 140 | Nemotron 3 Nano | 97.72% | 88.55% | 74.50% |
| 141 | Qwen 3.5 9B | 97.67% | 91.81% | 84.05% |
| 142 | Cydonia 24B V4.1 | 97.62% | 85.22% | 72.68% |
| 143 | Hermes 3 70B | 97.34% | 70.71% | 69.74% |
| 144 | WizardLM 2 8x22b | 95.09% | 79.28% | 71.45% |
| 145 | DeepSeek V3 (2025-03-24) | 94.58% | 76.88% | 79.93% |
| 146 | Llama 3.1 70B | 87.24% | 80.47% | 77.41% |