Posttrainbench — Epoch AI

Posttrainbench — Epoch AI is run by Epoch AI. It has ranked 12 models, of which the catalogue holds 11, scoring from 0.19 to 0.418.

measured by Epoch AI · the board itself

Place	Model	Metric	Score
1stof 12	Claude Fable 5	Average (%) (max)	0.418
2ndof 12	GPT-5.6 Sol	Average (%) (max)	0.362
3rdof 12	Claude Opus 5	Average (%)	0.35
4thof 12	Claude Opus 4.8	Average (%) (high)	0.338
6thof 12	Kimi K3	Average (%)	0.32
7thof 12	GLM 5.2	Average (%) (max)	0.317
8thof 12	Claude Opus 4.7	Average (%) (xhigh)	0.286
9thof 12	GPT-5.5	Average (%) (xhigh)	0.272
10thof 12	Grok 4.5	Average (%) (high)	0.234
11thof 12	Gemini 3.1 Pro Preview	Average (%)	0.22
12thof 12	GPT-5.4	Average (%) (high)	0.19

Place

Model

Metric

Score

1stof 12

Average (%) (max)

0.418

2ndof 12

Average (%) (max)

0.362