Enigma eval — Epoch AI

Enigma eval — Epoch AI is run by Epoch AI. It has ranked 46 models, of which the catalogue holds 35, scoring from 2.17 to 91.

measured by Epoch AI · the board itself

Place	Model	Metric	Score
1stof 46	Claude Fable 5	Accuracy (high)	39.28
2ndof 46	GPT-5.6 Sol	Accuracy (high)	37.12
3rdof 46	Gemini 3.1 Pro Preview	Accuracy	36.78
4thof 46	Gemini 3.5 Flash	Accuracy (high)	25.41
5thof 46	GPT-5.4 Pro	Accuracy	23.82
6thof 46	Claude Opus 4.8	Accuracy (xhigh)	23.51
7thof 46	GPT-5 Pro	Accuracy	18.75
8thof 46	gemini-3-pro-preview	Accuracy	18.24
9thof 46	GPT-5.4	Accuracy (xhigh)	15.96
10thof 46	o3	Accuracy (medium)	13.09
11thof 46	Claude Opus 4.5	Accuracy	11.91
13thof 46	GPT-5.1	Accuracy	11.23
14thof 46	GPT-5	Accuracy	10.47
15thof 46	GPT-5.2	Accuracy	10.39
16thof 46	o4-mini	Accuracy (high)	9.21
17thof 46	GPT-5 mini	Accuracy	8.193
18thof 46	Claude Opus 4.6	Accuracy (max)	7.6
19thof 46	Claude Opus 4.1	Accuracy	7.179
22ndof 46	o1-pro	Accuracy	6.14
23rdof 46	Claude Sonnet 4.5	Accuracy	5.997
24thof 46	o1	Accuracy	5.65
25thof 46	Claude Opus 4	Accuracy	5.57
26thof 46	Gemini 2.5 Pro Preview 06-05	Accuracy	5.57
27thof 46	Claude 3.7 Sonnet	Accuracy	4.23
29thof 46	Kimi K2.5	Accuracy	3.38
31stof 46	Claude Sonnet 4	Accuracy	3.12
32ndof 46	Gemini 3.1 Flash-Lite	Accuracy	3.04
33rdof 46	Gemini 2.5 Flash	Accuracy	2.7
34thof 46	Gemini 2.5 Pro Preview 05-06	Accuracy	2.36
36thof 46	GPT-4.1 fine-tuned	Accuracy	2.17
39thof 46	Claude 3.5 Sonnet	Accuracy	91
41stof 46	Claude 3 Opus	Accuracy	82
42ndof 46	GPT-4o (2024-11-20)	Accuracy	80
45thof 46	Llama-4-Maverick-17B-128E-Instruct	Accuracy	58
46thof 46	Llama-3.2-90B-Vision-Instruct	Accuracy	38

Place

Model

Metric

Score

1stof 46

Accuracy (high)

39.28

2ndof 46

Accuracy (high)

37.12