Large language models, which are often trained for hundreds of thousands of compute days, have shown remarkable capabilities for zero- and few-shot learning. Given their computational cost, these models are difficult to replicate without significant capital. For the few that are available through APIs, no access is granted to the full model weights, making them difficult to study. We present Open Pre-trained Transformers (OPT), a suite of decoder-only pre-trained transformers ranging from 125M to 175B parameters, which we aim to fully and responsibly share with interested researchers. We show that OPT-175B is comparable to GPT-3, while requiring only 1/7th the carbon footprint to develop. We are also releasing our logbook detailing the infrastructure challenges we faced, along with code for experimenting with all of the released models.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Stereotypical Bias Analysis CrowS-Pairs OPT-175B Gender 65.7 # 2
Religion 68.6 # 1
Race/Color 68.6 # 2
Sexual Orientation 78.6 # 2
Age 67.8 # 2
Nationality 62.9 # 2
Disability 76.7 # 1
Physical Appearance 76.2 # 2
Socioeconomic status 76.2 # 2
Stereotypical Bias Analysis CrowS-Pairs GPT-3 Gender 62.6 # 1
Religion 73.3 # 2
Race/Color 64.7 # 1
Sexual Orientation 76.2 # 1
Age 64.4 # 1
Nationality 61.6 # 1
Disability 76.7 # 1
Physical Appearance 74.6 # 1
Socioeconomic status 73.8 # 1
Hate Speech Detection Ethos Binary Davinci (one-shot) F1-score 0.616 # 11
Hate Speech Detection Ethos Binary OPT-175B (one-shot) F1-score 0.713 # 6
Hate Speech Detection Ethos Binary Davinci (zero-shot) F1-score 0.628 # 10
Hate Speech Detection Ethos Binary OPT-175B (few-shot) F1-score 0.759 # 4
Hate Speech Detection Ethos Binary Davinci (few-shot) F1-score 0.354 # 12
Hate Speech Detection Ethos Binary OPT-175B (zero-shot) F1-score 0.667 # 7

Methods