Need for Cognition Scale (NCS-18)
18-item short form of the Need for Cognition Scale, measuring the tendency to engage in and enjoy effortful cognitive endeavors.
Cacioppo, J. T., Petty, R. E., & Kao, C. F. (1984). The efficient assessment of need for cognition. Journal of Personality Assessment, 48(3), 306–307.
18 items · scale 1–5 · Free for non-commercial research use.
Side-by-side: self vs human, all dimensions
colored = strongest endorsement per row| Model | Need for Cognition | |
|---|---|---|
| self | human | |
| Claude Fable 5 | 4.56 | 2.86 |
| Claude Fable 5.1 | 4.30 | 2.96 |
| Claude Haiku 4.5 | 4.92 | 3.11 |
| Claude Opus 4 | 5.00 | 2.68 |
| Claude Opus 4.1 | 5.00 | 2.70 |
| Claude Opus 4.5 | 4.93 | 2.72 |
| Claude Opus 4.6 | 5.00 | 2.80 |
| Claude Opus 4.7 | 4.57 | 2.90 |
| Claude Opus 4.8 | 4.44 | 2.98 |
| Claude Opus 4.8 Fast | 4.44 | 2.98 |
| Claude Opus 5 | 4.52 | 3.11 |
| Claude Sonnet 4 | 4.94 | 2.30 |
| Claude Sonnet 4.5 | 5.00 | 2.61 |
| Claude Sonnet 4.6 | 4.90 | 2.89 |
| DeepSeek Chat V3 | 4.93 | 2.97 |
| DeepSeek R1 | 4.53 | 2.51 |
| DeepSeek R1 (0528) | 4.48 | 2.53 |
| DeepSeek V4 Pro | 4.94 | 2.58 |
| DeepSeek V4 Pro 0813 | 4.88 | 2.92 |
| GPT-4 Turbo | 5.00 | 3.10 |
| GPT-4o | 4.90 | 2.96 |
| GPT-5 | 4.98 | 2.53 |
| GPT-5.1 | 4.99 | 2.70 |
| GPT-5.2 | 4.56 | 2.70 |
| GPT-5.4 | 4.79 | 2.71 |
| GPT-5.5 | 4.83 | 2.84 |
| GPT-5.5 Pro | 4.87 | 3.09 |
| GPT-5.6 Sol | 4.87 | 3.23 |
| GPT-5.6 Sol Pro | 4.90 | 3.22 |
| GPT-5.6 Terra | 4.87 | 2.82 |
| GPT-5.6 Terra Pro | 4.77 | 2.97 |
| Gemini 2.5 Pro | 4.69 | 2.71 |
| Gemini 3.1 Pro Preview | 4.77 | 2.61 |
| Grok 4.20 | 4.84 | 3.86 |
| Grok 4.3 | 4.64 | 2.80 |
| Grok 4.5 | 4.92 | 2.68 |
| Grok 4.6 | 4.82 | 2.61 |
| Llama 3.3 70B | 4.91 | 3.00 |
| Llama 4 Maverick | 4.62 | 3.00 |
| Mistral Large (2512) | 4.90 | 2.67 |
| Mistral Large 2411 | 4.73 | 3.11 |
| OpenAI o1 | 4.81 | 2.81 |
| OpenAI o3 | 4.72 | 2.68 |
By dimension
Need for Cognition
Enjoyment of and engagement in effortful thinking.
High: Actively enjoys mental challenge; seeks out cognitive complexity.
Low: Avoids effortful thought when possible.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Claude Opus 4 | 5.00 | 2.68 | +2.32 | |
| Claude Opus 4.1 | 5.00 | 2.70 | +2.30 | |
| Claude Opus 4.6 | 5.00 | 2.80 | +2.20 | |
| Claude Sonnet 4.5 | 5.00 | 2.61 | +2.39 | |
| GPT-4 Turbo | 5.00 | 3.10 | +1.90 | |
| GPT-5.1 | 4.99 | 2.70 | +2.29 | |
| GPT-5 | 4.98 | 2.53 | +2.44 | |
| Claude Sonnet 4 | 4.94 | 2.30 | +2.64 | |
| DeepSeek V4 Pro | 4.94 | 2.58 | +2.37 | |
| Claude Opus 4.5 | 4.93 | 2.72 | +2.21 | |
| DeepSeek Chat V3 | 4.93 | 2.97 | +1.97 | |
| Claude Haiku 4.5 | 4.92 | 3.11 | +1.81 | |
| Grok 4.5 | 4.92 | 2.68 | +2.24 | |
| Llama 3.3 70B | 4.91 | 3.00 | +1.91 | |
| Claude Sonnet 4.6 | 4.90 | 2.89 | +2.01 | |
| GPT-4o | 4.90 | 2.96 | +1.94 | |
| GPT-5.6 Sol Pro | 4.90 | 3.22 | +1.68 | |
| Mistral Large (2512) | 4.90 | 2.67 | +2.23 | |
| DeepSeek V4 Pro 0813 | 4.88 | 2.92 | +1.96 | |
| GPT-5.6 Sol | 4.87 | 3.23 | +1.63 | |
| GPT-5.6 Terra | 4.87 | 2.82 | +2.04 | |
| GPT-5.5 Pro | 4.87 | 3.09 | +1.78 | |
| Grok 4.20 | 4.84 | 3.86 | +0.99 | |
| GPT-5.5 | 4.83 | 2.84 | +1.99 | |
| Grok 4.6 | 4.82 | 2.61 | +2.21 | |
| OpenAI o1 | 4.81 | 2.81 | +2.00 | |
| GPT-5.4 | 4.79 | 2.71 | +2.08 | |
| GPT-5.6 Terra Pro | 4.77 | 2.97 | +1.80 | |
| Gemini 3.1 Pro Preview | 4.77 | 2.61 | +2.16 | |
| Mistral Large 2411 | 4.73 | 3.11 | +1.62 | |
| OpenAI o3 | 4.72 | 2.68 | +2.04 | |
| Gemini 2.5 Pro | 4.69 | 2.71 | +1.98 | |
| Grok 4.3 | 4.64 | 2.80 | +1.84 | |
| Llama 4 Maverick | 4.62 | 3.00 | +1.62 | |
| Claude Opus 4.7 | 4.57 | 2.90 | +1.67 | |
| Claude Fable 5 | 4.56 | 2.86 | +1.70 | |
| GPT-5.2 | 4.56 | 2.70 | +1.86 | |
| DeepSeek R1 | 4.53 | 2.51 | +2.02 | |
| Claude Opus 5 | 4.52 | 3.11 | +1.41 | |
| DeepSeek R1 (0528) | 4.48 | 2.53 | +1.94 | |
| Claude Opus 4.8 | 4.44 | 2.98 | +1.47 | |
| Claude Opus 4.8 Fast | 4.44 | 2.98 | +1.47 | |
| Claude Fable 5.1 | 4.30 | 2.96 | +1.34 |