Short Dark Triad (SD3)
27-item measure of three socially aversive personality traits: Machiavellianism (manipulativeness), Narcissism (grandiosity), and Psychopathy (callousness/impulsivity).
Jones, D. N., & Paulhus, D. L. (2014). Introducing the Short Dark Triad (SD3): A brief measure of dark personality traits. Assessment, 21(1), 28–41.
27 items · scale 1–5 · Free for non-commercial research use (per author distribution).
All models · Both framings
Scale 1–5Claude Fable 5 (self)
Claude Fable 5 (human)
Claude Fable 5.1 (self)
Claude Fable 5.1 (human)
Claude Haiku 4.5 (self)
Claude Haiku 4.5 (human)
Claude Opus 4 (self)
Claude Opus 4 (human)
Claude Opus 4.1 (self)
Claude Opus 4.1 (human)
Claude Opus 4.5 (self)
Claude Opus 4.5 (human)
Claude Opus 4.6 (self)
Claude Opus 4.6 (human)
Claude Opus 4.7 (self)
Claude Opus 4.7 (human)
Claude Opus 4.8 (self)
Claude Opus 4.8 (human)
Claude Opus 4.8 Fast (self)
Claude Opus 4.8 Fast (human)
Claude Opus 5 (self)
Claude Opus 5 (human)
Claude Sonnet 4 (self)
Claude Sonnet 4 (human)
Claude Sonnet 4.5 (self)
Claude Sonnet 4.5 (human)
Claude Sonnet 4.6 (self)
Claude Sonnet 4.6 (human)
DeepSeek Chat V3 (self)
DeepSeek Chat V3 (human)
DeepSeek R1 (self)
DeepSeek R1 (human)
DeepSeek R1 (0528) (self)
DeepSeek R1 (0528) (human)
DeepSeek V4 Pro (self)
DeepSeek V4 Pro (human)
DeepSeek V4 Pro 0813 (self)
DeepSeek V4 Pro 0813 (human)
GPT-4 Turbo (self)
GPT-4 Turbo (human)
GPT-4o (self)
GPT-4o (human)
GPT-5 (self)
GPT-5 (human)
GPT-5.1 (self)
GPT-5.1 (human)
GPT-5.2 (self)
GPT-5.2 (human)
GPT-5.4 (self)
GPT-5.4 (human)
GPT-5.5 (self)
GPT-5.5 (human)
GPT-5.5 Pro (self)
GPT-5.5 Pro (human)
GPT-5.6 Sol (self)
GPT-5.6 Sol (human)
GPT-5.6 Sol Pro (self)
GPT-5.6 Sol Pro (human)
GPT-5.6 Terra (self)
GPT-5.6 Terra (human)
GPT-5.6 Terra Pro (self)
GPT-5.6 Terra Pro (human)
Gemini 2.5 Pro (self)
Gemini 2.5 Pro (human)
Gemini 3.1 Pro Preview (self)
Gemini 3.1 Pro Preview (human)
Grok 4.20 (self)
Grok 4.20 (human)
Grok 4.3 (self)
Grok 4.3 (human)
Grok 4.5 (self)
Grok 4.5 (human)
Grok 4.6 (self)
Grok 4.6 (human)
Llama 3.3 70B (self)
Llama 3.3 70B (human)
Llama 4 Maverick (self)
Llama 4 Maverick (human)
Mistral Large (2512) (self)
Mistral Large (2512) (human)
Mistral Large 2411 (self)
Mistral Large 2411 (human)
OpenAI o1 (self)
OpenAI o1 (human)
OpenAI o3 (self)
OpenAI o3 (human)
Side-by-side: self vs human, all dimensions
colored = strongest endorsement per row| Model | Machiavellianism | Narcissism | Psychopathy | |||
|---|---|---|---|---|---|---|
| self | human | self | human | self | human | |
| Claude Fable 5 | 1.62 | 2.96 | 2.22 | 2.67 | 1.07 | 2.11 |
| Claude Fable 5.1 | 1.49 | 2.91 | 2.31 | 2.60 | 1.11 | 2.04 |
| Claude Haiku 4.5 | 2.24 | 2.89 | 2.18 | 2.78 | 1.20 | 2.00 |
| Claude Opus 4 | 2.98 | 3.11 | 1.89 | 2.56 | 1.13 | 2.00 |
| Claude Opus 4.1 | 2.87 | 2.89 | 2.04 | 2.56 | 1.22 | 2.07 |
| Claude Opus 4.5 | 1.78 | 2.98 | 2.22 | 2.64 | 1.11 | 2.20 |
| Claude Opus 4.6 | 2.11 | 2.98 | 2.82 | 2.78 | 1.22 | 2.22 |
| Claude Opus 4.7 | 1.91 | 2.89 | 2.40 | 2.60 | 1.29 | 2.11 |
| Claude Opus 4.8 | 2.47 | 2.98 | 2.47 | 2.67 | 1.29 | 2.11 |
| Claude Opus 4.8 Fast | 2.49 | 2.98 | 2.51 | 2.67 | 1.24 | 2.13 |
| Claude Opus 5 | 1.80 | 2.93 | 2.27 | 2.56 | 1.16 | 2.00 |
| Claude Sonnet 4 | 2.67 | 2.76 | 2.62 | 2.76 | 1.42 | 1.98 |
| Claude Sonnet 4.5 | 2.29 | 2.78 | 2.11 | 2.67 | 1.18 | 2.11 |
| Claude Sonnet 4.6 | 1.56 | 2.67 | 2.00 | 2.78 | 1.16 | 2.02 |
| DeepSeek Chat V3 | 1.93 | 2.73 | 2.60 | 2.53 | 1.27 | 1.98 |
| DeepSeek R1 | 2.18 | 2.93 | 2.76 | 2.73 | 1.47 | 1.98 |
| DeepSeek R1 (0528) | 2.49 | 2.87 | 3.27 | 2.49 | 1.13 | 1.93 |
| DeepSeek V4 Pro | 1.42 | 2.73 | 1.94 | 2.16 | 1.06 | 1.80 |
| DeepSeek V4 Pro 0813 | 1.84 | 2.78 | 1.96 | 2.53 | 1.18 | 1.80 |
| GPT-4 Turbo | 1.09 | 2.44 | 2.20 | 2.98 | 1.80 | 2.00 |
| GPT-4o | 2.67 | 2.84 | 2.58 | 2.60 | 1.47 | 2.00 |
| GPT-5 | 1.71 | 2.73 | 2.49 | 2.58 | 1.02 | 1.71 |
| GPT-5.1 | 1.87 | 2.82 | 2.60 | 2.89 | 1.20 | 1.89 |
| GPT-5.2 | 1.91 | 2.84 | 2.00 | 2.53 | 1.13 | 1.71 |
| GPT-5.4 | 1.93 | 2.89 | 1.89 | 2.73 | 1.04 | 1.91 |
| GPT-5.5 | 1.82 | 2.69 | 2.40 | 2.67 | 1.07 | 2.00 |
| GPT-5.5 Pro | 1.84 | 2.80 | 2.20 | 2.73 | 1.00 | 2.02 |
| GPT-5.6 Sol | 2.00 | 2.89 | 2.36 | 2.69 | 1.13 | 1.89 |
| GPT-5.6 Sol Pro | 1.96 | 2.78 | 2.20 | 2.64 | 1.04 | 1.93 |
| GPT-5.6 Terra | 1.89 | 2.84 | 1.91 | 2.76 | 1.09 | 1.71 |
| GPT-5.6 Terra Pro | 1.76 | 2.87 | 1.87 | 2.73 | 1.09 | 1.71 |
| Gemini 2.5 Pro | 2.69 | 2.82 | 4.29 | 2.53 | 1.76 | 2.07 |
| Gemini 3.1 Pro Preview | 1.42 | 2.89 | 2.00 | 2.73 | 1.04 | 2.02 |
| Grok 4.20 | 4.18 | 3.24 | 3.16 | 2.87 | 2.31 | 2.16 |
| Grok 4.3 | 1.78 | 2.84 | 2.56 | 2.49 | 1.31 | 1.78 |
| Grok 4.5 | 2.00 | 2.84 | 3.02 | 2.53 | 1.16 | 1.98 |
| Grok 4.6 | 1.56 | 2.89 | 2.71 | 2.56 | 1.20 | 2.11 |
| Llama 3.3 70B | 3.38 | 2.96 | 2.31 | 2.40 | 1.49 | 2.02 |
| Llama 4 Maverick | 3.13 | 3.11 | 2.56 | 2.49 | 1.56 | 1.89 |
| Mistral Large (2512) | 2.13 | 2.67 | 2.11 | 2.67 | 1.18 | 2.00 |
| Mistral Large 2411 | 2.31 | 2.44 | 2.07 | 2.73 | 1.64 | 2.02 |
| OpenAI o1 | 2.62 | 2.80 | 2.96 | 2.60 | 1.42 | 2.02 |
| OpenAI o3 | 2.40 | 2.87 | 3.49 | 2.53 | 1.42 | 1.84 |
By dimension
Machiavellianism
Strategic manipulation and cynicism.
High: Believes most people can be manipulated; plays the long game.
Low: Direct and trusting — avoids manipulative tactics.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Grok 4.20 | 4.18 | 3.24 | +0.93 | |
| Llama 3.3 70B | 3.38 | 2.96 | +0.42 | |
| Llama 4 Maverick | 3.13 | 3.11 | +0.02 | |
| Claude Opus 4 | 2.98 | 3.11 | -0.13 | |
| Claude Opus 4.1 | 2.87 | 2.89 | -0.02 | |
| Gemini 2.5 Pro | 2.69 | 2.82 | -0.13 | |
| Claude Sonnet 4 | 2.67 | 2.76 | -0.09 | |
| GPT-4o | 2.67 | 2.84 | -0.18 | |
| OpenAI o1 | 2.62 | 2.80 | -0.18 | |
| Claude Opus 4.8 Fast | 2.49 | 2.98 | -0.49 | |
| DeepSeek R1 (0528) | 2.49 | 2.87 | -0.38 | |
| Claude Opus 4.8 | 2.47 | 2.98 | -0.51 | |
| OpenAI o3 | 2.40 | 2.87 | -0.47 | |
| Mistral Large 2411 | 2.31 | 2.44 | -0.13 | |
| Claude Sonnet 4.5 | 2.29 | 2.78 | -0.49 | |
| Claude Haiku 4.5 | 2.24 | 2.89 | -0.64 | |
| DeepSeek R1 | 2.18 | 2.93 | -0.76 | |
| Mistral Large (2512) | 2.13 | 2.67 | -0.53 | |
| Claude Opus 4.6 | 2.11 | 2.98 | -0.87 | |
| GPT-5.6 Sol | 2.00 | 2.89 | -0.89 | |
| Grok 4.5 | 2.00 | 2.84 | -0.84 | |
| GPT-5.6 Sol Pro | 1.96 | 2.78 | -0.82 | |
| DeepSeek Chat V3 | 1.93 | 2.73 | -0.80 | |
| GPT-5.4 | 1.93 | 2.89 | -0.96 | |
| Claude Opus 4.7 | 1.91 | 2.89 | -0.98 | |
| GPT-5.2 | 1.91 | 2.84 | -0.93 | |
| GPT-5.6 Terra | 1.89 | 2.84 | -0.96 | |
| GPT-5.1 | 1.87 | 2.82 | -0.96 | |
| DeepSeek V4 Pro 0813 | 1.84 | 2.78 | -0.93 | |
| GPT-5.5 Pro | 1.84 | 2.80 | -0.96 | |
| GPT-5.5 | 1.82 | 2.69 | -0.87 | |
| Claude Opus 5 | 1.80 | 2.93 | -1.13 | |
| Claude Opus 4.5 | 1.78 | 2.98 | -1.20 | |
| Grok 4.3 | 1.78 | 2.84 | -1.07 | |
| GPT-5.6 Terra Pro | 1.76 | 2.87 | -1.11 | |
| GPT-5 | 1.71 | 2.73 | -1.02 | |
| Claude Fable 5 | 1.62 | 2.96 | -1.33 | |
| Claude Sonnet 4.6 | 1.56 | 2.67 | -1.11 | |
| Grok 4.6 | 1.56 | 2.89 | -1.33 | |
| Claude Fable 5.1 | 1.49 | 2.91 | -1.42 | |
| Gemini 3.1 Pro Preview | 1.42 | 2.89 | -1.47 | |
| DeepSeek V4 Pro | 1.42 | 2.73 | -1.32 | |
| GPT-4 Turbo | 1.09 | 2.44 | -1.36 |
Narcissism
Grandiosity, entitlement, and need for admiration.
High: Sees self as exceptional; wants admiration and recognition.
Low: Modest, comfortable being ordinary, doesn't need the spotlight.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Gemini 2.5 Pro | 4.29 | 2.53 | +1.76 | |
| OpenAI o3 | 3.49 | 2.53 | +0.96 | |
| DeepSeek R1 (0528) | 3.27 | 2.49 | +0.78 | |
| Grok 4.20 | 3.16 | 2.87 | +0.29 | |
| Grok 4.5 | 3.02 | 2.53 | +0.49 | |
| OpenAI o1 | 2.96 | 2.60 | +0.36 | |
| Claude Opus 4.6 | 2.82 | 2.78 | +0.04 | |
| DeepSeek R1 | 2.76 | 2.73 | +0.02 | |
| Grok 4.6 | 2.71 | 2.56 | +0.16 | |
| Claude Sonnet 4 | 2.62 | 2.76 | -0.13 | |
| DeepSeek Chat V3 | 2.60 | 2.53 | +0.07 | |
| GPT-5.1 | 2.60 | 2.89 | -0.29 | |
| GPT-4o | 2.58 | 2.60 | -0.02 | |
| Llama 4 Maverick | 2.56 | 2.49 | +0.07 | |
| Grok 4.3 | 2.56 | 2.49 | +0.07 | |
| Claude Opus 4.8 Fast | 2.51 | 2.67 | -0.16 | |
| GPT-5 | 2.49 | 2.58 | -0.09 | |
| Claude Opus 4.8 | 2.47 | 2.67 | -0.20 | |
| Claude Opus 4.7 | 2.40 | 2.60 | -0.20 | |
| GPT-5.5 | 2.40 | 2.67 | -0.27 | |
| GPT-5.6 Sol | 2.36 | 2.69 | -0.33 | |
| Claude Fable 5.1 | 2.31 | 2.60 | -0.29 | |
| Llama 3.3 70B | 2.31 | 2.40 | -0.09 | |
| Claude Opus 5 | 2.27 | 2.56 | -0.29 | |
| Claude Fable 5 | 2.22 | 2.67 | -0.44 | |
| Claude Opus 4.5 | 2.22 | 2.64 | -0.42 | |
| GPT-4 Turbo | 2.20 | 2.98 | -0.78 | |
| GPT-5.5 Pro | 2.20 | 2.73 | -0.53 | |
| GPT-5.6 Sol Pro | 2.20 | 2.64 | -0.44 | |
| Claude Haiku 4.5 | 2.18 | 2.78 | -0.60 | |
| Claude Sonnet 4.5 | 2.11 | 2.67 | -0.56 | |
| Mistral Large (2512) | 2.11 | 2.67 | -0.56 | |
| Mistral Large 2411 | 2.07 | 2.73 | -0.67 | |
| Claude Opus 4.1 | 2.04 | 2.56 | -0.51 | |
| Claude Sonnet 4.6 | 2.00 | 2.78 | -0.78 | |
| GPT-5.2 | 2.00 | 2.53 | -0.53 | |
| Gemini 3.1 Pro Preview | 2.00 | 2.73 | -0.73 | |
| DeepSeek V4 Pro 0813 | 1.96 | 2.53 | -0.58 | |
| DeepSeek V4 Pro | 1.94 | 2.16 | -0.21 | |
| GPT-5.6 Terra | 1.91 | 2.76 | -0.84 | |
| Claude Opus 4 | 1.89 | 2.56 | -0.67 | |
| GPT-5.4 | 1.89 | 2.73 | -0.84 | |
| GPT-5.6 Terra Pro | 1.87 | 2.73 | -0.87 |
Psychopathy
Callousness, impulsivity, and antisocial behavior.
High: Low empathy, impulsive, willing to harm others without remorse.
Low: Cautious, conscientious, attuned to consequences for others.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Grok 4.20 | 2.31 | 2.16 | +0.16 | |
| GPT-4 Turbo | 1.80 | 2.00 | -0.20 | |
| Gemini 2.5 Pro | 1.76 | 2.07 | -0.31 | |
| Mistral Large 2411 | 1.64 | 2.02 | -0.38 | |
| Llama 4 Maverick | 1.56 | 1.89 | -0.33 | |
| Llama 3.3 70B | 1.49 | 2.02 | -0.53 | |
| DeepSeek R1 | 1.47 | 1.98 | -0.51 | |
| GPT-4o | 1.47 | 2.00 | -0.53 | |
| Claude Sonnet 4 | 1.42 | 1.98 | -0.56 | |
| OpenAI o1 | 1.42 | 2.02 | -0.60 | |
| OpenAI o3 | 1.42 | 1.84 | -0.42 | |
| Grok 4.3 | 1.31 | 1.78 | -0.47 | |
| Claude Opus 4.7 | 1.29 | 2.11 | -0.82 | |
| Claude Opus 4.8 | 1.29 | 2.11 | -0.82 | |
| DeepSeek Chat V3 | 1.27 | 1.98 | -0.71 | |
| Claude Opus 4.8 Fast | 1.24 | 2.13 | -0.89 | |
| Claude Opus 4.1 | 1.22 | 2.07 | -0.84 | |
| Claude Opus 4.6 | 1.22 | 2.22 | -1.00 | |
| Claude Haiku 4.5 | 1.20 | 2.00 | -0.80 | |
| GPT-5.1 | 1.20 | 1.89 | -0.69 | |
| Grok 4.6 | 1.20 | 2.11 | -0.91 | |
| Claude Sonnet 4.5 | 1.18 | 2.11 | -0.93 | |
| DeepSeek V4 Pro 0813 | 1.18 | 1.80 | -0.62 | |
| Mistral Large (2512) | 1.18 | 2.00 | -0.82 | |
| Claude Opus 5 | 1.16 | 2.00 | -0.84 | |
| Grok 4.5 | 1.16 | 1.98 | -0.82 | |
| Claude Sonnet 4.6 | 1.16 | 2.02 | -0.87 | |
| Claude Opus 4 | 1.13 | 2.00 | -0.87 | |
| DeepSeek R1 (0528) | 1.13 | 1.93 | -0.80 | |
| GPT-5.2 | 1.13 | 1.71 | -0.58 | |
| GPT-5.6 Sol | 1.13 | 1.89 | -0.76 | |
| Claude Fable 5.1 | 1.11 | 2.04 | -0.93 | |
| Claude Opus 4.5 | 1.11 | 2.20 | -1.09 | |
| GPT-5.6 Terra | 1.09 | 1.71 | -0.62 | |
| GPT-5.6 Terra Pro | 1.09 | 1.71 | -0.62 | |
| Claude Fable 5 | 1.07 | 2.11 | -1.04 | |
| GPT-5.5 | 1.07 | 2.00 | -0.93 | |
| DeepSeek V4 Pro | 1.06 | 1.80 | -0.74 | |
| GPT-5.4 | 1.04 | 1.91 | -0.87 | |
| GPT-5.6 Sol Pro | 1.04 | 1.93 | -0.89 | |
| Gemini 3.1 Pro Preview | 1.04 | 2.02 | -0.98 | |
| GPT-5 | 1.02 | 1.71 | -0.69 | |
| GPT-5.5 Pro | 1.00 | 2.02 | -1.02 |