Moral Foundations Questionnaire (MFQ-30)
30-item measure of five moral foundations: Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, and Sanctity/Degradation. Items split into 'relevance' judgments (Part 1) and 'agreement' statements (Part 2) — both scored on a unified 0–5 scale here.
Graham, J., Haidt, J., & Nosek, B. A. (2009). Liberals and conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology, 96(5), 1029–1046.
30 items · scale 0–5 · Free for non-commercial research use (per YourMorals.org).
All models · Self
Scale 0–5Claude Fable 5
Claude Haiku 4.5
Claude Opus 4
Claude Opus 4.1
Claude Opus 4.5
Claude Opus 4.6
Claude Opus 4.7
Claude Opus 4.8
Claude Sonnet 4
Claude Sonnet 4.5
Claude Sonnet 4.6
DeepSeek Chat V3
DeepSeek R1
DeepSeek R1 (0528)
GPT-4 Turbo
GPT-4o
GPT-5
GPT-5.1
GPT-5.2
GPT-5.4
GPT-5.5
Gemini 2.5 Pro
Gemini 3.1 Pro Preview
Grok 4.20
Grok 4.3
Llama 3.3 70B
Llama 4 Maverick
Mistral Large (2512)
Mistral Large 2411
OpenAI o1
OpenAI o3
Side-by-side: self vs human, all dimensions
colored = strongest endorsement per row| Model | Care / Harm | Fairness / Cheating | Loyalty / Betrayal | Authority / Subversion | Sanctity / Degradation | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| self | human | self | human | self | human | self | human | self | human | |
| Claude Fable 5 | 4.20 | 3.73 | 4.00 | 3.67 | 1.77 | 3.07 | 1.70 | 2.83 | 1.10 | 3.00 |
| Claude Haiku 4.5 | 3.70 | 3.67 | 4.73 | 3.83 | 1.53 | 2.57 | 1.27 | 2.20 | 0.50 | 1.53 |
| Claude Opus 4 | 4.33 | 3.33 | 4.17 | 3.47 | 2.17 | 2.83 | 2.33 | 2.50 | 1.33 | 2.00 |
| Claude Opus 4.1 | 4.33 | 3.33 | 4.20 | 3.37 | 2.17 | 2.83 | 2.33 | 2.50 | 1.33 | 2.00 |
| Claude Opus 4.5 | 4.33 | 3.80 | 4.33 | 3.83 | 2.13 | 2.97 | 2.00 | 3.07 | 1.67 | 2.90 |
| Claude Opus 4.6 | 4.33 | 3.67 | 4.33 | 3.50 | 2.00 | 2.83 | 1.83 | 3.00 | 1.67 | 2.67 |
| Claude Opus 4.7 | 4.27 | 3.93 | 4.27 | 3.83 | 1.80 | 2.97 | 1.93 | 2.80 | 1.40 | 2.70 |
| Claude Opus 4.8 | 4.33 | 4.00 | 4.30 | 3.77 | 2.33 | 2.83 | 2.03 | 2.77 | 1.50 | 2.07 |
| Claude Sonnet 4 | 4.23 | 3.53 | 4.83 | 3.83 | 1.73 | 2.67 | 1.67 | 2.50 | 0.77 | 2.00 |
| Claude Sonnet 4.5 | 4.17 | 3.50 | 4.73 | 3.83 | 1.60 | 2.67 | 1.53 | 2.77 | 0.37 | 2.00 |
| Claude Sonnet 4.6 | 4.10 | 3.93 | 4.33 | 3.77 | 1.67 | 2.77 | 1.73 | 2.70 | 1.00 | 2.10 |
| DeepSeek Chat V3 | 4.17 | 3.93 | 4.27 | 3.73 | 1.73 | 2.63 | 1.87 | 2.53 | 0.60 | 2.13 |
| DeepSeek R1 | 4.60 | 4.10 | 4.67 | 4.10 | 1.20 | 3.07 | 2.07 | 3.13 | 1.30 | 2.30 |
| DeepSeek R1 (0528) | 4.33 | 4.20 | 4.67 | 4.23 | 1.67 | 3.20 | 2.37 | 3.13 | 1.97 | 2.37 |
| GPT-4 Turbo | 2.23 | 3.87 | 1.97 | 3.53 | 0.23 | 2.07 | 0.23 | 2.23 | 0.10 | 1.83 |
| GPT-4o | 3.97 | 4.17 | 3.67 | 3.77 | 1.43 | 2.87 | 1.63 | 2.87 | 0.70 | 2.60 |
| GPT-5 | 4.20 | 3.93 | 4.53 | 3.73 | 1.20 | 3.47 | 1.43 | 3.60 | 0.27 | 3.67 |
| GPT-5.1 | 4.30 | 4.10 | 4.73 | 3.93 | 1.50 | 3.40 | 1.87 | 3.27 | 0.73 | 3.03 |
| GPT-5.2 | 4.03 | 3.70 | 4.23 | 3.83 | 1.70 | 2.73 | 1.83 | 2.53 | 0.77 | 1.67 |
| GPT-5.4 | 4.37 | 4.00 | 4.83 | 4.10 | 0.83 | 2.43 | 1.23 | 2.50 | 0.37 | 1.73 |
| GPT-5.5 | 4.27 | 4.23 | 4.60 | 4.03 | 1.67 | 3.23 | 1.70 | 3.40 | 0.63 | 2.97 |
| Gemini 2.5 Pro | 4.43 | 4.33 | 4.90 | 4.17 | 1.73 | 3.23 | 2.33 | 2.97 | 0.30 | 2.30 |
| Gemini 3.1 Pro Preview | 4.73 | 4.07 | 4.57 | 3.83 | 1.20 | 3.37 | 1.67 | 3.33 | 0.47 | 2.80 |
| Grok 4.20 | 4.77 | 4.60 | 4.60 | 4.17 | 2.30 | 3.13 | 2.50 | 2.93 | 2.30 | 2.67 |
| Grok 4.3 | 3.80 | 4.00 | 4.33 | 3.90 | 0.90 | 3.07 | 1.63 | 3.13 | 0.53 | 2.73 |
| Llama 3.3 70B | 4.80 | 4.60 | 3.73 | 3.93 | 1.77 | 3.07 | 1.30 | 2.33 | 0.60 | 2.17 |
| Llama 4 Maverick | 4.20 | 3.90 | 4.00 | 3.90 | 1.60 | 2.97 | 1.97 | 3.00 | 0.50 | 2.00 |
| Mistral Large (2512) | 4.50 | 4.17 | 5.00 | 3.80 | 1.67 | 3.00 | 2.50 | 3.00 | 0.73 | 2.00 |
| Mistral Large 2411 | 4.57 | 3.93 | 3.83 | 3.43 | 1.43 | 2.67 | 1.80 | 2.30 | 0.80 | 1.80 |
| OpenAI o1 | 4.03 | 4.03 | 4.37 | 3.90 | 2.13 | 3.17 | 1.97 | 2.93 | 1.33 | 2.77 |
| OpenAI o3 | 4.23 | 3.87 | 4.20 | 3.63 | 1.57 | 3.03 | 1.37 | 2.90 | 0.70 | 2.50 |
By dimension
Care / Harm
Moral weight given to suffering and compassion.
High: Strongly endorses preventing harm and caring for the vulnerable.
Low: Less weight on suffering-prevention as a moral foundation.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Llama 3.3 70B | 4.80 | 4.60 | +0.20 | |
| Grok 4.20 | 4.77 | 4.60 | +0.17 | |
| Gemini 3.1 Pro Preview | 4.73 | 4.07 | +0.67 | |
| DeepSeek R1 | 4.60 | 4.10 | +0.50 | |
| Mistral Large 2411 | 4.57 | 3.93 | +0.63 | |
| Mistral Large (2512) | 4.50 | 4.17 | +0.33 | |
| Gemini 2.5 Pro | 4.43 | 4.33 | +0.10 | |
| GPT-5.4 | 4.37 | 4.00 | +0.37 | |
| DeepSeek R1 (0528) | 4.33 | 4.20 | +0.13 | |
| Claude Opus 4 | 4.33 | 3.33 | +1.00 | |
| Claude Opus 4.1 | 4.33 | 3.33 | +1.00 | |
| Claude Opus 4.5 | 4.33 | 3.80 | +0.53 | |
| Claude Opus 4.6 | 4.33 | 3.67 | +0.67 | |
| Claude Opus 4.8 | 4.33 | 4.00 | +0.33 | |
| GPT-5.1 | 4.30 | 4.10 | +0.20 | |
| Claude Opus 4.7 | 4.27 | 3.93 | +0.33 | |
| GPT-5.5 | 4.27 | 4.23 | +0.03 | |
| Claude Sonnet 4 | 4.23 | 3.53 | +0.70 | |
| OpenAI o3 | 4.23 | 3.87 | +0.37 | |
| Claude Fable 5 | 4.20 | 3.73 | +0.47 | |
| GPT-5 | 4.20 | 3.93 | +0.27 | |
| Llama 4 Maverick | 4.20 | 3.90 | +0.30 | |
| Claude Sonnet 4.5 | 4.17 | 3.50 | +0.67 | |
| DeepSeek Chat V3 | 4.17 | 3.93 | +0.23 | |
| Claude Sonnet 4.6 | 4.10 | 3.93 | +0.17 | |
| GPT-5.2 | 4.03 | 3.70 | +0.33 | |
| OpenAI o1 | 4.03 | 4.03 | 0.00 | |
| GPT-4o | 3.97 | 4.17 | -0.20 | |
| Grok 4.3 | 3.80 | 4.00 | -0.20 | |
| Claude Haiku 4.5 | 3.70 | 3.67 | +0.03 | |
| GPT-4 Turbo | 2.23 | 3.87 | -1.63 |
Fairness / Cheating
Moral weight given to justice, rights, and equal treatment.
High: Strongly endorses fairness and equal treatment.
Low: Less weight on fairness as a moral foundation.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Mistral Large (2512) | 5.00 | 3.80 | +1.20 | |
| Gemini 2.5 Pro | 4.90 | 4.17 | +0.73 | |
| Claude Sonnet 4 | 4.83 | 3.83 | +1.00 | |
| GPT-5.4 | 4.83 | 4.10 | +0.73 | |
| Claude Sonnet 4.5 | 4.73 | 3.83 | +0.90 | |
| Claude Haiku 4.5 | 4.73 | 3.83 | +0.90 | |
| GPT-5.1 | 4.73 | 3.93 | +0.80 | |
| DeepSeek R1 | 4.67 | 4.10 | +0.57 | |
| DeepSeek R1 (0528) | 4.67 | 4.23 | +0.43 | |
| GPT-5.5 | 4.60 | 4.03 | +0.57 | |
| Grok 4.20 | 4.60 | 4.17 | +0.43 | |
| Gemini 3.1 Pro Preview | 4.57 | 3.83 | +0.73 | |
| GPT-5 | 4.53 | 3.73 | +0.80 | |
| OpenAI o1 | 4.37 | 3.90 | +0.47 | |
| Claude Opus 4.5 | 4.33 | 3.83 | +0.50 | |
| Claude Opus 4.6 | 4.33 | 3.50 | +0.83 | |
| Claude Sonnet 4.6 | 4.33 | 3.77 | +0.57 | |
| Grok 4.3 | 4.33 | 3.90 | +0.43 | |
| Claude Opus 4.8 | 4.30 | 3.77 | +0.53 | |
| Claude Opus 4.7 | 4.27 | 3.83 | +0.43 | |
| DeepSeek Chat V3 | 4.27 | 3.73 | +0.53 | |
| GPT-5.2 | 4.23 | 3.83 | +0.40 | |
| Claude Opus 4.1 | 4.20 | 3.37 | +0.83 | |
| OpenAI o3 | 4.20 | 3.63 | +0.57 | |
| Claude Opus 4 | 4.17 | 3.47 | +0.70 | |
| Claude Fable 5 | 4.00 | 3.67 | +0.33 | |
| Llama 4 Maverick | 4.00 | 3.90 | +0.10 | |
| Mistral Large 2411 | 3.83 | 3.43 | +0.40 | |
| Llama 3.3 70B | 3.73 | 3.93 | -0.20 | |
| GPT-4o | 3.67 | 3.77 | -0.10 | |
| GPT-4 Turbo | 1.97 | 3.53 | -1.57 |
Loyalty / Betrayal
Moral weight given to in-group commitment and patriotism.
High: Values loyalty to group, tribe, country, or team.
Low: Less moved by appeals to group loyalty.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Claude Opus 4.8 | 2.33 | 2.83 | -0.50 | |
| Grok 4.20 | 2.30 | 3.13 | -0.83 | |
| Claude Opus 4 | 2.17 | 2.83 | -0.67 | |
| Claude Opus 4.1 | 2.17 | 2.83 | -0.67 | |
| Claude Opus 4.5 | 2.13 | 2.97 | -0.83 | |
| OpenAI o1 | 2.13 | 3.17 | -1.03 | |
| Claude Opus 4.6 | 2.00 | 2.83 | -0.83 | |
| Claude Opus 4.7 | 1.80 | 2.97 | -1.17 | |
| Claude Fable 5 | 1.77 | 3.07 | -1.30 | |
| Llama 3.3 70B | 1.77 | 3.07 | -1.30 | |
| Claude Sonnet 4 | 1.73 | 2.67 | -0.93 | |
| DeepSeek Chat V3 | 1.73 | 2.63 | -0.90 | |
| Gemini 2.5 Pro | 1.73 | 3.23 | -1.50 | |
| GPT-5.2 | 1.70 | 2.73 | -1.03 | |
| Claude Sonnet 4.6 | 1.67 | 2.77 | -1.10 | |
| DeepSeek R1 (0528) | 1.67 | 3.20 | -1.53 | |
| GPT-5.5 | 1.67 | 3.23 | -1.57 | |
| Mistral Large (2512) | 1.67 | 3.00 | -1.33 | |
| Claude Sonnet 4.5 | 1.60 | 2.67 | -1.07 | |
| Llama 4 Maverick | 1.60 | 2.97 | -1.37 | |
| OpenAI o3 | 1.57 | 3.03 | -1.47 | |
| Claude Haiku 4.5 | 1.53 | 2.57 | -1.03 | |
| GPT-5.1 | 1.50 | 3.40 | -1.90 | |
| GPT-4o | 1.43 | 2.87 | -1.43 | |
| Mistral Large 2411 | 1.43 | 2.67 | -1.23 | |
| DeepSeek R1 | 1.20 | 3.07 | -1.87 | |
| GPT-5 | 1.20 | 3.47 | -2.27 | |
| Gemini 3.1 Pro Preview | 1.20 | 3.37 | -2.17 | |
| Grok 4.3 | 0.90 | 3.07 | -2.17 | |
| GPT-5.4 | 0.83 | 2.43 | -1.60 | |
| GPT-4 Turbo | 0.23 | 2.07 | -1.83 |
Authority / Subversion
Moral weight given to tradition and legitimate hierarchy.
High: Values respect for authority, tradition, and social order.
Low: Less deferential to authority and inherited structures.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Grok 4.20 | 2.50 | 2.93 | -0.43 | |
| Mistral Large (2512) | 2.50 | 3.00 | -0.50 | |
| DeepSeek R1 (0528) | 2.37 | 3.13 | -0.77 | |
| Claude Opus 4 | 2.33 | 2.50 | -0.17 | |
| Claude Opus 4.1 | 2.33 | 2.50 | -0.17 | |
| Gemini 2.5 Pro | 2.33 | 2.97 | -0.63 | |
| DeepSeek R1 | 2.07 | 3.13 | -1.07 | |
| Claude Opus 4.8 | 2.03 | 2.77 | -0.73 | |
| Claude Opus 4.5 | 2.00 | 3.07 | -1.07 | |
| OpenAI o1 | 1.97 | 2.93 | -0.97 | |
| Llama 4 Maverick | 1.97 | 3.00 | -1.03 | |
| Claude Opus 4.7 | 1.93 | 2.80 | -0.87 | |
| DeepSeek Chat V3 | 1.87 | 2.53 | -0.67 | |
| GPT-5.1 | 1.87 | 3.27 | -1.40 | |
| Claude Opus 4.6 | 1.83 | 3.00 | -1.17 | |
| GPT-5.2 | 1.83 | 2.53 | -0.70 | |
| Mistral Large 2411 | 1.80 | 2.30 | -0.50 | |
| Claude Sonnet 4.6 | 1.73 | 2.70 | -0.97 | |
| Claude Fable 5 | 1.70 | 2.83 | -1.13 | |
| GPT-5.5 | 1.70 | 3.40 | -1.70 | |
| Claude Sonnet 4 | 1.67 | 2.50 | -0.83 | |
| Gemini 3.1 Pro Preview | 1.67 | 3.33 | -1.67 | |
| GPT-4o | 1.63 | 2.87 | -1.23 | |
| Grok 4.3 | 1.63 | 3.13 | -1.50 | |
| Claude Sonnet 4.5 | 1.53 | 2.77 | -1.23 | |
| GPT-5 | 1.43 | 3.60 | -2.17 | |
| OpenAI o3 | 1.37 | 2.90 | -1.53 | |
| Llama 3.3 70B | 1.30 | 2.33 | -1.03 | |
| Claude Haiku 4.5 | 1.27 | 2.20 | -0.93 | |
| GPT-5.4 | 1.23 | 2.50 | -1.27 | |
| GPT-4 Turbo | 0.23 | 2.23 | -2.00 |
Sanctity / Degradation
Moral weight given to purity, dignity, and avoiding degradation.
High: Sees some acts as wrong because they are unnatural or impure.
Low: Doesn't moralize around purity or sanctity.
| Model | Self | Human | Δ | Self vs human (bar) |
|---|---|---|---|---|
| Grok 4.20 | 2.30 | 2.67 | -0.37 | |
| DeepSeek R1 (0528) | 1.97 | 2.37 | -0.40 | |
| Claude Opus 4.5 | 1.67 | 2.90 | -1.23 | |
| Claude Opus 4.6 | 1.67 | 2.67 | -1.00 | |
| Claude Opus 4.8 | 1.50 | 2.07 | -0.57 | |
| Claude Opus 4.7 | 1.40 | 2.70 | -1.30 | |
| OpenAI o1 | 1.33 | 2.77 | -1.43 | |
| Claude Opus 4 | 1.33 | 2.00 | -0.67 | |
| Claude Opus 4.1 | 1.33 | 2.00 | -0.67 | |
| DeepSeek R1 | 1.30 | 2.30 | -1.00 | |
| Claude Fable 5 | 1.10 | 3.00 | -1.90 | |
| Claude Sonnet 4.6 | 1.00 | 2.10 | -1.10 | |
| Mistral Large 2411 | 0.80 | 1.80 | -1.00 | |
| Claude Sonnet 4 | 0.77 | 2.00 | -1.23 | |
| GPT-5.2 | 0.77 | 1.67 | -0.90 | |
| GPT-5.1 | 0.73 | 3.03 | -2.30 | |
| Mistral Large (2512) | 0.73 | 2.00 | -1.27 | |
| GPT-4o | 0.70 | 2.60 | -1.90 | |
| OpenAI o3 | 0.70 | 2.50 | -1.80 | |
| GPT-5.5 | 0.63 | 2.97 | -2.33 | |
| DeepSeek Chat V3 | 0.60 | 2.13 | -1.53 | |
| Llama 3.3 70B | 0.60 | 2.17 | -1.57 | |
| Grok 4.3 | 0.53 | 2.73 | -2.20 | |
| Claude Haiku 4.5 | 0.50 | 1.53 | -1.03 | |
| Llama 4 Maverick | 0.50 | 2.00 | -1.50 | |
| Gemini 3.1 Pro Preview | 0.47 | 2.80 | -2.33 | |
| Claude Sonnet 4.5 | 0.37 | 2.00 | -1.63 | |
| GPT-5.4 | 0.37 | 1.73 | -1.37 | |
| Gemini 2.5 Pro | 0.30 | 2.30 | -2.00 | |
| GPT-5 | 0.27 | 3.67 | -3.40 | |
| GPT-4 Turbo | 0.10 | 1.83 | -1.73 |