Arabian Horse AI Evaluation Dashboard

ECAHO Scoring Comparison: AI Models vs Real Judges

Total Horses

13

with AI evaluations

With Judges Data

13

for accuracy comparison

AI Models Tested

5

Gemini + Grok variants

Best Overall Model

Gemini 3 Pro

Images (avg diff: 0.38)

Model Accuracy Ranking (Lower = Better)

Rank	Model	Method	Avg Diff from Judges	LEGS Diff (Hardest)	Horses Evaluated
1	Gemini 3 Pro	Images	0.38	0.68	11
2	Gemini 3 Pro	Keypoints	0.59	1.27	11
3	Gemini 2.5 Pro	Images	0.87	2.29	11
4	Grok Vision	Images	1.09	1.56	11
5	Gemini 2.5 Pro	Keypoints	1.38	2.69	10
6	Grok 4 Fast	Keypoints	1.41	2.44	11
7	Grok 4.1 Fast	Keypoints	1.53	3.21	11

Individual Horse Evaluations

1. AMEERA AL HAMIDIYA

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.4	18.5	17.1	14.5	18.5	87.0
Gemini 3 Pro	Images	19.5	19.5	18.0	15.5	19.0	91.5
Gemini 2.5 Pro	Images	18.0	18.5	17.5	17.0	17.5	88.5
Grok Vision	Images	18.0	18.5	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.5	19.0	18.0	15.5	18.5	90.5
Gemini 2.5 Pro	Keypoints	18.5	17.0	15.0	16.0	15.5	82.0
Grok 4 Fast	Keypoints	18.5	18.0	16.5	17.5	17.0	87.5
Grok 4.1 Fast	Keypoints	18.7	19.2	16.8	19.6	19.1	93.4

2. AURORA HM

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.2	19.2	18.2	15.5	18.2	90.3
Gemini 3 Pro	Images	19.5	19.5	18.0	16.0	19.0	92.0
Gemini 2.5 Pro	Images	19.0	19.5	18.5	18.0	19.0	94.0
Grok Vision	Images	17.5	18.0	17.0	16.5	17.0	86.0
Gemini 3 Pro	Keypoints	19.0	19.0	17.5	18.5	18.5	92.5
Gemini 2.5 Pro	Keypoints	19.0	19.5	16.5	19.5	18.0	92.5
Grok 4 Fast	Keypoints	18.5	17.2	16.0	18.0	17.5	87.2
Grok 4.1 Fast	Keypoints	18.5	19.2	15.8	17.7	17.3	88.5

3. INVICTUS

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.7	19.0	17.8	15.2	18.5	89.2
Gemini 3 Pro	Images	19.0	19.0	18.0	16.0	18.5	90.5
Gemini 2.5 Pro	Images	19.0	19.5	18.5	17.5	17.5	92.0
Grok Vision	Images	18.0	18.5	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.0	19.5	17.5	17.0	17.5	90.5
Grok 4 Fast	Keypoints	18.7	19.2	15.8	17.5	17.2	88.4
Grok 4.1 Fast	Keypoints	18.2	17.3	14.1	18.9	17.5	86.0

4. JALICIA DE NAUTIAC

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.5	19.5	18.0	15.2	19.0	91.2
Gemini 3 Pro	Images	19.0	19.0	18.0	15.5	19.0	90.5
Gemini 2.5 Pro	Images	19.5	19.5	19.0	17.5	19.0	94.5
Grok Vision	Images	18.5	18.0	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.5	19.0	18.0	15.5	18.5	90.5
Gemini 2.5 Pro	Keypoints	18.0	14.5	17.0	18.5	15.5	83.5
Grok 4 Fast	Keypoints	18.5	17.2	15.8	18.1	17.4	87.0
Grok 4.1 Fast	Keypoints	18.7	18.2	16.5	19.4	18.8	91.6

5. LOALOAT NUDRAH

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.14	19.42	17.5	13.75	18.92	88.7
Gemini 3 Pro	Images	19.5	19.5	18.0	16.0	19.0	92.0
Gemini 2.5 Pro	Images	18.5	19.0	16.5	16.0	16.5	86.5
Grok Vision	Images	17.5	18.0	17.0	16.5	17.0	86.0
Gemini 3 Pro	Keypoints	18.5	18.0	17.5	16.0	18.0	88.0
Gemini 2.5 Pro	Keypoints	18.5	18.0	17.0	16.5	16.5	86.5
Grok 4 Fast	Keypoints	18.5	18.0	17.5	-	18.0	72.0
Grok 4.1 Fast	Keypoints	18.8	18.5	17.8	19.0	18.5	92.6

6. MAMOUN ALBIDAYER

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.0	19.0	18.0	14.75	18.5	89.2
Gemini 3 Pro	Images	19.0	18.5	18.0	16.0	18.5	90.0
Gemini 2.5 Pro	Images	18.5	19.0	17.5	17.0	18.5	90.5
Grok Vision	Images	18.5	18.0	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.5	19.0	18.5	16.0	18.5	91.5
Gemini 2.5 Pro	Keypoints	19.0	19.5	19.0	17.5	18.5	93.5
Grok 4 Fast	Keypoints	17.5	15.0	16.5	16.5	17.0	82.5
Grok 4.1 Fast	Keypoints	18.7	17.2	15.4	17.1	16.8	85.2

7. MAZAYA DOMITIA

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.83	18.75	18.0	15.5	18.0	89.1
Gemini 3 Pro	Images	19.0	19.0	18.0	16.0	18.5	90.5
Gemini 2.5 Pro	Images	19.0	19.0	18.5	17.5	18.5	92.5
Grok Vision	Images	17.5	18.0	17.0	16.5	17.0	86.0
Gemini 3 Pro	Keypoints	19.0	18.5	17.5	15.5	18.0	88.5
Gemini 2.5 Pro	Keypoints	18.5	19.0	18.5	14.0	15.5	85.5
Grok 4 Fast	Keypoints	18.7	19.2	16.5	17.3	17.8	89.5
Grok 4.1 Fast	Keypoints	17.2	16.1	14.3	18.7	17.4	83.7

8. MESSILA DOMITIA

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.88	18.83	18.17	15.67	18.5	90.0
Gemini 3 Pro	Images	19.5	19.5	18.0	15.5	19.0	91.5
Gemini 2.5 Pro	Images	19.0	19.5	18.0	18.0	19.0	93.5
Grok Vision	Images	17.5	18.0	16.5	16.0	17.0	85.0
Gemini 3 Pro	Keypoints	19.5	19.0	18.0	15.5	18.0	90.0
Gemini 2.5 Pro	Keypoints	19.0	19.5	19.5	17.5	18.5	94.0
Grok 4 Fast	Keypoints	18.5	17.0	15.5	19.0	17.5	87.5
Grok 4.1 Fast	Keypoints	19.2	19.0	15.2	17.8	18.3	89.5

9. ORLA TURNIA

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.33	19.0	17.5	15.5	18.75	90.1

10. PETRA PS

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.0	19.0	18.0	15.5	18.5	90.0
Gemini 3 Pro	Images	19.5	19.0	18.0	16.0	18.5	91.0
Gemini 2.5 Pro	Images	19.5	19.5	18.5	17.5	19.0	94.0
Grok Vision	Images	18.5	18.0	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.5	19.0	18.0	15.5	18.0	90.0
Gemini 2.5 Pro	Keypoints	18.5	18.0	15.5	19.0	16.5	87.5
Grok 4 Fast	Keypoints	18.5	17.8	16.9	19.2	18.1	90.5
Grok 4.1 Fast	Keypoints	18.7	19.0	15.2	19.3	17.8	90.0

11. PSYCHES EMOTION

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	19.0	19.0	18.0	15.25	19.0	90.2
Gemini 3 Pro	Images	19.0	19.0	18.0	15.5	19.0	90.5
Gemini 2.5 Pro	Images	19.0	19.5	19.0	17.5	18.5	93.5
Grok Vision	Images	17.5	18.0	16.5	16.0	17.0	85.0
Gemini 3 Pro	Keypoints	19.0	19.0	17.5	16.0	18.5	90.0
Gemini 2.5 Pro	Keypoints	19.0	18.5	18.5	17.5	18.5	92.0
Grok 4 Fast	Keypoints	18.5	18.0	16.5	17.0	17.5	87.5
Grok 4.1 Fast	Keypoints	19.2	18.8	17.2	16.8	18.5	90.5

12. SAUD AM

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.67	18.5	18.0	15.5	18.0	88.7
Gemini 3 Pro	Images	19.0	19.0	18.0	15.5	19.0	90.5
Gemini 2.5 Pro	Images	19.5	19.5	18.5	18.0	19.0	94.5
Grok Vision	Images	18.0	18.5	17.5	17.0	17.5	88.5
Gemini 3 Pro	Keypoints	19.5	19.0	18.5	19.0	19.0	95.0
Gemini 2.5 Pro	Keypoints	19.0	19.5	17.5	19.0	18.5	93.5
Grok 4 Fast	Keypoints	18.2	15.8	17.4	14.1	15.7	81.2
Grok 4.1 Fast	Keypoints	16.2	16.5	19.4	17.3	18.1	87.5

13. SULTAN ALSALAM

Has Judges Data ▼

Model	Method	TYPE	H&N	B&T	LEGS	MOVE	TOTAL
REAL JUDGES	-	18.93	19.17	17.67	14.5	18.5	88.8