Piqa — Epoch AI

Piqa — Epoch AI is run by Epoch AI. It has ranked 113 models, of which the catalogue holds 31, scoring from 65.8 to 88.7 on Score.

measured by Epoch AI · the board itself

Place	Model	Metric	Score
1stof 113	GPT-4o-mini (2024-07-18)	Score	88.7
2ndof 113	Phi-3.5-MoE-instruct	Score	88.6
4thof 113	Llama 3.1 405B	Score	85.9
5thof 113	falcon-180B	Score	84.9
8thof 113	DeepSeek-V2	Score	83.9
9thof 113	gemma-2-9b	Score	83.7
12thof 113	StableBeluga2	Score	83.3
14thof 113	Mistral-7B-v0.1	Score	83
15thof 113	falcon-40b	Score	83
17thof 113	Llama-2-70b-hf	Score	82.8
33rdof 113	Mistral-7B-Instruct-v0.2	Score	82.2
35thof 113	Mistral-7B-Instruct-v0.1	Score	82.2
48thof 113	gemma-7b	Score	81.2
49thof 113	Llama 3.1 8B Instruct	Score	81.2
51stof 113	Phi-3.5-mini-instruct	Score	81
56thof 113	Llama-2-13b	Score	80.8
68thof 113	falcon-7b	Score	80.3
69thof 113	llama-13b	Score	80.1
72ndof 113	Qwen-14B	Score	79.9
74thof 113	llama-7b	Score	79.8
82ndof 113	Llama-2-7b	Score	78.8
91stof 113	Mistral 7B	Score	77.9
95thof 113	gemma-2b	Score	77.3
99thof 113	gpt-neox-20b	Score	76.7
100thof 113	Baichuan-7B	Score	76.2
102ndof 113	opt-1.3b	Score	75.7
103rdof 113	xgen-7b-8k-base	Score	75.5
104thof 113	gpt-j-6b	Score	75.4
109thof 113	gpt-neo-2.7B	Score	72.9
110thof 113	gpt2-xl	Score	70.5
113thof 113	stablelm-tuned-alpha-7b	Score	65.8

Place

Model

Metric

Score

1stof 113

Score

88.7

2ndof 113

Score

88.6