Adversarial nli — Epoch AI
Adversarial nli — Epoch AI is run by Epoch AI. It has ranked 15 models, of which the catalogue holds 8, scoring from 42.5 to 58.1 on Score.
measured by Epoch AI · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 1stof 15 | Phi-3-small-8k-instruct | Score | 58.1 |
| 2ndof 15 | GPT-3.5 Turbo (1106) | Score | 58.1 |
| 3rdof 15 | Llama 3 8B Instruct | Score | 57.3 |
| 4thof 15 | Phi-3-medium-128k-instruct | Score | 55.8 |
| 6thof 15 | Phi-3-mini-4k-instruct | Score | 52.8 |
| 7thof 15 | gemma-7b | Score | 48.7 |
| 8thof 15 | Mistral-7B-v0.1 | Score | 47.1 |
| 9thof 15 | phi-2 | Score | 42.5 |
Read from the board on 2026-09-16