Share your thoughts, 1 month free Claude Pro on usSee more

Commonsense Reasoning on OBQA (first-token accuracy)

91.4First-Token Accuracy

Phi-4-14B

Updated 12d ago

Evaluation Results

Method	Links
Phi-4-14B 2025.05		91.4
Phi-4-14B 2025.05		90
Gemma-2-9B 2025.05		89.8
Gemma-2-9B 2025.05		87.6
Ministral-8B 2025.05		85.2
Llama-3.1-8B 2025.05		84.8
Qwen-2-7B 2025.05		84.8
Qwen-2-7B 2025.05		84.2
Phi-4-14B 2025.05		84
Qwen-2-7B 2025.05		83.8
Ministral-8B 2025.05		82.4
Mistral-Nemo-12B 2025.05		80.8
Mistral-Nemo-12B 2025.05		80.6
Llama-3.1-8B 2025.05		80.4
Ministral-8B 2025.05		76.4
Zephyr-7B 2025.05		73
Llama-3.1-8B 2025.05		71.2
Gemma-7B 2025.05		71.2
Zephyr-7B 2025.05		70.8
Mistral-Nemo-12B 2025.05		69.8
Zephyr-7B 2025.05		67.6
Gemma-7B 2025.05		63.8
Gemma-7B 2025.05		40.6
Gemma-2-9B 2025.05		38.8