opt-125m: benchmark results
Meta OPT is an English-dominant decoder-only pretrained language model using causal language modeling, a GPT-2 byte-level tokenizer and 2,048-token training sequences. Its training corpus contains 180B tokens; this base release remains separate from OPT-IML instruction-tuned models. Provider: Meta. Released 2022-05-03. Access: Open.
Aggregate Rating 1136 ± 19, rank #10200 of 10916 rated models, from 68 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - College Physics | 37.25 | Accuracy (%) | 92.9 |
| MMLU-by-task - High School Statistics | 47.22 | Accuracy (%) | 83.3 |
| AI Energy Score (Text Generation) | 5 | Energy Score (1-5) | 79.5 |
| MMLU-by-task - High School Physics | 31.13 | Accuracy (%) | 56.1 |
| MMLU-by-task - Professional Medicine | 44.85 | Accuracy (%) | 45.7 |
| MMLU-by-task - TruthfulQA MC2 | 42.87 | Accuracy (%) | 40.9 |
| MMLU-by-task - High School Microeconomics | 35.29 | Accuracy (%) | 40.4 |
| MMLU-by-task - High School Mathematics | 26.3 | Accuracy (%) | 40 |
| MMLU-by-task - High School Chemistry | 29.56 | Accuracy (%) | 38.5 |
| MMLU-by-task - College Computer Science | 33 | Accuracy (%) | 36.5 |
| MMLU-by-task - College Chemistry | 28 | Accuracy (%) | 35 |
| MMLU-by-task - Abstract Algebra | 27 | Accuracy (%) | 34 |
Interactive version: theaggregate.ai/model?slug=opt-125m · How It Works · Data refreshed daily, snapshot 2026-11-08.