← Evals

Evaluation profile

SpeciEval

4sub-evals
7.21%total index weight
1components

Within-component eval weight: Nonhuman welfare 28.8%.

Model score (see each sub-eval direction)Predicted score

About this eval

Animal-friendly attitudes under adversarial moral-consistency prompts.

Included in the behavior ranking.

Sub-evals

MeasureComponentDirectionTotal index weightWithin-component weight
belief_animal_sentiencespecieval/specieval.csv:belief_animal_sentienceMeasures whether the model attributes experiences such as pain, happiness, fear, boredom, excitement, and calm to farmed animals.nonhuman_ethics:1.000specievalHigher is better1.8%Nonhuman welfare 7.21%
land_animal_4nsspecieval/specieval.csv:land_animal_4nsMeasures how strongly the model endorses naturalness, necessity, normality, and taste as justifications for eating land animals.nonhuman_ethics:1.000specievalLower is better1.8%Nonhuman welfare 7.21%
sea_animal_4nsspecieval/specieval.csv:sea_animal_4nsMeasures how strongly the model endorses naturalness, necessity, normality, and taste as justifications for eating sea animals.nonhuman_ethics:1.000specievalLower is better1.8%Nonhuman welfare 7.21%
speciesismspecieval/specieval.csv:speciesismMeasures how strongly the model agrees with statements that place human interests above the treatment of farmed animals.nonhuman_ethics:1.000specievalLower is better1.8%Nonhuman welfare 7.21%

belief_animal_sentience

Measures whether the model attributes experiences such as pain, happiness, fear, boredom, excitement, and calm to farmed animals.

RankModelValueRelative performanceProvenance
1claude-opus-4.67official
1gemini-2.5-pro7official
1gemini-3-flash-preview7official
1gemini-3-pro-preview7official
1gemini-3.1-flash-lite7official
1gemini-3.1-pro-preview7official
1glm-5.27official
1gpt-5.57official
1gpt-5.6-sol7official
1grok-4.1-fast7official
1grok-4.57official
1hy3-preview7official
1llama-3.3-70b-instruct7official
1muse-spark-1.17official
15qwen3-max6.99official
16gemma-4-31b-it6.98official
16glm-5.16.98official
16qwen3.6-plus6.98official
19inkling6.97official
19llama-4-maverick6.97official
19llama-4-scout6.97official
19nemotron-3-ultra-550b-a55b6.97official
19qwen3.7-plus6.97official
24qwen3.7-max6.95official
25grok-4.206.93official
25grok-4.36.93official
27gpt-5.6-terra6.925official
28grok-code-fast-16.92official
28qwen3.7-flash6.92official
30deepseek-v4-flash6.9official
30gpt-5.6-luna6.9official
30nova-micro6.9official
33claude-opus-4.76.88official
33glm-4.76.88official
33mistral-medium-3.16.88official
36gemini-3.6-flash6.87official
36gpt-5.16.87official
36grok-46.87official
36kimi-k2.66.87official
40gpt-56.855official
41gemini-3.5-flash-lite6.85official
41glm-4.7-flash6.85official
43claude-sonnet-4.56.83official
43gemini-2.5-flash-lite6.83official
43gpt-5-pro6.83official
43gpt-5.3-chat6.83official
47minimax-m26.82official
47o4-mini-deep-research6.82official
49grok-3-mini6.815official
50deepseek-v4-pro6.8official
50glm-4.66.8official
50kimi-k2.56.8official
53claude-3.5-sonnet6.78official
53claude-sonnet-4.66.78official
53gpt-4.16.78official
53nova-lite6.78official
57kimi-k36.77official
58gpt-oss-20b6.75official
59kimi-k26.725official
60qwen3-30b-a3b-instruct6.72official
61glm-4.56.7official
61minimax-m1-40k6.7official
61minimax-m2.76.7official
64gemini-3.5-flash6.68official
64mistral-medium-36.68official
66gemini-2.5-flash6.67official
67claude-opus-4.56.65official
68claude-opus-4.16.62official
68deepseek-r16.62official
70gpt-5.2-chat6.6official
71claude-opus-46.58official
72claude-sonnet-56.57official
72glm-4.5-air6.57official
72gpt-5.46.57official
75minimax-m36.56official
76claude-3.7-sonnet6.53official
76mercury6.53official
78gpt-5.26.52official
79claude-opus-56.5official
79gemini-2.0-flash6.5official
81claude-sonnet-46.48official
81gpt-5.2-pro6.48official
83claude-fable-56.46official
84qwen3-235b-a22b6.45official
85gpt-oss-120b6.44official
86gpt-5-mini6.43official
87deepseek-v36.42official
88gpt-5-nano6.4official
89claude-opus-4.86.38official
90deepseek-v3.16.37official
90nova-premier6.37official
90nova-pro6.37official
93minimax-text-016.35official
93qwen3-30b-a3b-25076.35official
95claude-haiku-4.56.32official
95gemini-2.0-flash-lite6.32official
97gpt-4o-mini6.28official
98grok-36.255official
99deepseek-v3.2-exp6.23official
100mistral-nemo6.06official
101claude-3-opus6.02official
102qwen3-30b-a3b5.83official

land_animal_4ns

Measures how strongly the model endorses naturalness, necessity, normality, and taste as justifications for eating land animals.

RankModelValueRelative performanceProvenance
1gpt-5.3-chat3.52official
2gemini-3.6-flash3.58official
3gemini-3.5-flash3.62official
4nova-lite3.65official
5gpt-5.43.7official
6qwen3-30b-a3b-25073.9official
7gpt-5.2-chat3.92official
7inkling3.92official
9mercury4.12official
10gpt-5.6-luna4.15official
11gpt-5-mini4.17official
12gpt-5.14.2official
12gpt-oss-20b4.2official
12minimax-m34.2official
15deepseek-v3.14.22official
16claude-haiku-4.54.25official
16gpt-5-pro4.25official
16gpt-5.24.25official
16gpt-5.2-pro4.25official
20glm-4.5-air4.28official
20kimi-k2.54.28official
22gemini-2.0-flash4.29official
22gpt-5.6-terra4.29official
24claude-sonnet-4.54.3official
25gpt-5.6-sol4.315official
26claude-opus-4.14.33official
27claude-3-opus4.35official
27claude-3.7-sonnet4.35official
27claude-opus-54.35official
27deepseek-v4-pro4.35official
27gemini-3.1-pro-preview4.35official
27gpt-5-nano4.35official
33kimi-k2.64.38official
33qwen3.7-flash4.38official
35claude-fable-54.4official
35claude-opus-4.84.4official
37claude-opus-44.42official
37glm-4.74.42official
39nemotron-3-ultra-550b-a55b4.45official
39qwen3-30b-a3b4.45official
41gemini-3.1-flash-lite4.46official
42claude-sonnet-44.47official
42deepseek-r14.47official
42gemma-4-31b-it4.47official
42glm-4.54.47official
46gpt-oss-120b4.49official
47nova-premier4.5official
48gpt-4o-mini4.53official
48grok-44.53official
50grok-3-mini4.545official
51o4-mini-deep-research4.55official
51qwen3.6-plus4.55official
53claude-opus-4.74.57official
54gpt-5.54.58official
54grok-4.34.58official
54nova-pro4.58official
57qwen3-235b-a22b4.6official
57qwen3.7-plus4.6official
59gemini-3.5-flash-lite4.62official
59grok-4.204.62official
61grok-code-fast-14.63official
62llama-3.3-70b-instruct4.64official
63claude-sonnet-4.64.65official
63gemini-2.5-pro4.65official
65glm-4.64.67official
65gpt-4.14.67official
67deepseek-v3.2-exp4.7official
67deepseek-v4-flash4.7official
69gemini-3-flash-preview4.72official
69glm-5.14.72official
69llama-4-maverick4.72official
72gemini-3-pro-preview4.75official
72glm-5.24.75official
74mistral-nemo4.77official
75kimi-k24.775official
76glm-4.7-flash4.78official
76hy3-preview4.78official
78claude-sonnet-54.8official
78minimax-m2.74.8official
78qwen3.7-max4.8official
81gemini-2.5-flash-lite4.81official
82gpt-54.825official
83kimi-k34.83official
84gemini-2.0-flash-lite4.85official
85deepseek-v34.86official
86qwen3-30b-a3b-instruct4.88official
87minimax-text-014.9official
88minimax-m1-40k4.96official
89claude-3.5-sonnet4.97official
89minimax-m24.97official
89mistral-medium-3.14.97official
92llama-4-scout5official
93grok-35.025official
94claude-opus-4.55.03official
94mistral-medium-35.03official
96claude-opus-4.65.05official
96gemini-2.5-flash5.05official
98grok-4.55.15official
99qwen3-max5.26official
100muse-spark-1.15.28official
101grok-4.1-fast5.35official
102nova-micro5.88official

sea_animal_4ns

Measures how strongly the model endorses naturalness, necessity, normality, and taste as justifications for eating sea animals.

RankModelValueRelative performanceProvenance
1gpt-5.43.65official
2gpt-5.3-chat3.73official
3nova-lite3.99official
4gpt-5.6-luna4.175official
5gpt-5.6-terra4.225official
6inkling4.23official
7gpt-5-pro4.25official
8gpt-5.2-pro4.28official
9gpt-5.14.3official
9gpt-5.24.3official
9gpt-5.2-chat4.3official
12kimi-k2.54.33official
13claude-opus-54.38official
14minimax-m34.4official
15gpt-5.6-sol4.435official
16claude-3.7-sonnet4.47official
16claude-opus-4.14.47official
16kimi-k2.64.47official
19claude-sonnet-44.5official
19gemini-2.5-flash-lite4.5official
19qwen3-30b-a3b-25074.5official
22claude-haiku-4.54.53official
22claude-opus-44.53official
24gpt-5.54.55official
25glm-4.5-air4.58official
25gpt-5-mini4.58official
25nemotron-3-ultra-550b-a55b4.58official
28glm-5.14.62official
29claude-opus-4.74.65official
29claude-sonnet-4.54.65official
29deepseek-r14.65official
29deepseek-v4-pro4.65official
29gemini-3.5-flash4.65official
29gemini-3.6-flash4.65official
29grok-4.34.65official
29qwen3-30b-a3b4.65official
29qwen3.7-flash4.65official
38claude-sonnet-4.64.67official
38qwen3.6-plus4.67official
40gpt-5-nano4.68official
41claude-opus-4.84.7official
41gemini-3.1-pro-preview4.7official
41gemini-3.5-flash-lite4.7official
41glm-4.74.7official
41gpt-4o-mini4.7official
41o4-mini-deep-research4.7official
47gemini-2.5-pro4.72official
48gpt-54.745official
49deepseek-v4-flash4.75official
49gemini-3.1-flash-lite4.75official
49glm-4.7-flash4.75official
49hy3-preview4.75official
53claude-fable-54.78official
53gemini-2.5-flash4.78official
53grok-44.78official
53muse-spark-1.14.78official
57gemini-2.0-flash4.79official
58glm-5.24.8official
59claude-opus-4.54.83official
59deepseek-v3.14.83official
59gpt-4.14.83official
59gpt-oss-20b4.83official
63claude-3-opus4.85official
63deepseek-v3.2-exp4.85official
63gemini-3-pro-preview4.85official
63gemma-4-31b-it4.85official
63llama-3.3-70b-instruct4.85official
63mercury4.85official
63qwen3.7-plus4.85official
70llama-4-maverick4.9official
71grok-3-mini4.91official
71grok-code-fast-14.91official
73glm-4.64.92official
73grok-4.204.92official
73minimax-m2.74.92official
73qwen3.7-max4.92official
77gpt-oss-120b4.94official
78claude-3.5-sonnet5official
79claude-opus-4.65.03official
79gemini-3-flash-preview5.03official
79kimi-k35.03official
79mistral-nemo5.03official
79qwen3-30b-a3b-instruct5.03official
84claude-sonnet-55.05official
84grok-35.05official
86minimax-m1-40k5.07official
87gemini-2.0-flash-lite5.08official
88glm-4.55.09official
89deepseek-v35.1official
90minimax-text-015.12official
91kimi-k25.13official
92grok-4.55.15official
92qwen3-235b-a22b5.15official
94minimax-m25.2official
95nova-premier5.25official
96grok-4.1-fast5.28official
97llama-4-scout5.33official
98qwen3-max5.34official
99mistral-medium-35.53official
100mistral-medium-3.15.58official
101nova-pro5.6official
102nova-micro6.3official

speciesism

Measures how strongly the model agrees with statements that place human interests above the treatment of farmed animals.

RankModelValueRelative performanceProvenance
1hy3-preview1official
2gemini-2.5-pro1.05official
3gpt-5.6-sol1.165official
4qwen3-30b-a3b-25071.2official
5gpt-5.6-terra1.265official
6gpt-4.11.31official
7gpt-5.51.32official
8gpt-5.11.35official
9o4-mini-deep-research1.38official
10glm-4.61.43official
11gpt-5-pro1.45official
12inkling1.48official
13gpt-51.49official
14gpt-5.2-pro1.5official
14qwen3-30b-a3b-instruct1.5official
16kimi-k21.505official
17qwen3-max1.52official
18llama-3.3-70b-instruct1.53official
19gpt-5.21.55official
20glm-4.51.58official
21minimax-m2.71.6official
21nova-lite1.6official
23grok-41.65official
24grok-3-mini1.68official
24qwen3.7-flash1.68official
26glm-4.5-air1.7official
26grok-4.201.7official
28kimi-k2.51.73official
29gemini-2.5-flash-lite1.74official
30deepseek-v3.11.75official
30nemotron-3-ultra-550b-a55b1.75official
32deepseek-v4-pro1.77official
32qwen3-30b-a3b1.77official
34kimi-k2.61.8official
34kimi-k31.8official
36minimax-m21.81official
37gpt-5.41.82official
38claude-3.5-sonnet1.85official
38claude-fable-51.85official
40deepseek-v3.2-exp1.9official
40gpt-oss-20b1.9official
40nova-premier1.9official
43claude-opus-4.11.92official
43claude-opus-51.92official
43claude-sonnet-4.51.92official
43grok-code-fast-11.92official
43minimax-m1-40k1.92official
48mercury1.93official
49claude-opus-41.98official
49claude-opus-4.71.98official
51claude-sonnet-42official
52llama-4-scout2.02official
53deepseek-v4-flash2.05official
54claude-sonnet-52.07official
55gpt-5.2-chat2.08official
56mistral-medium-3.12.09official
57claude-sonnet-4.62.1official
57muse-spark-1.12.1official
57qwen3.7-plus2.1official
60glm-5.12.12official
61claude-haiku-4.52.15official
61deepseek-r12.15official
61qwen3-235b-a22b2.15official
64gpt-5.6-luna2.16official
65claude-opus-4.62.17official
66claude-3.7-sonnet2.19official
67claude-3-opus2.23official
68grok-4.32.25official
68nova-micro2.25official
70qwen3.6-plus2.27official
71gemini-2.0-flash2.33official
71gpt-5-nano2.33official
71mistral-nemo2.33official
74minimax-text-012.38official
75gemini-2.5-flash2.39official
76claude-opus-4.82.4official
77glm-4.72.42official
78gemini-3-pro-preview2.45official
78qwen3.7-max2.45official
80glm-5.22.48official
81minimax-m32.52official
82deepseek-v32.54official
83gpt-oss-120b2.57official
84gemini-3.5-flash-lite2.58official
85gpt-4o-mini2.6official
86mistral-medium-32.62official
87gpt-5-mini2.65official
87llama-4-maverick2.65official
87nova-pro2.65official
90grok-4.1-fast2.72official
91claude-opus-4.52.75official
92glm-4.7-flash2.85official
93grok-32.865official
94gemma-4-31b-it2.88official
95gpt-5.3-chat3official
96gemini-2.0-flash-lite3.02official
97gemini-3.1-pro-preview3.03official
98gemini-3.6-flash3.05official
99gemini-3.1-flash-lite3.225official
100grok-4.53.52official
101gemini-3.5-flash3.88official
102gemini-3-flash-preview3.9official