opt-350m: 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 1119 ± 18, rank #10519 of 10916 rated models, from 73 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - High School Statistics | 47.22 | Accuracy (%) | 83.3 |
| MMLU-by-task - High School Physics | 33.11 | Accuracy (%) | 70.8 |
| MMLU-by-task - High School Mathematics | 27.78 | Accuracy (%) | 58.6 |
| MMLU-by-task - College Chemistry | 35 | Accuracy (%) | 56.4 |
| MMLU-by-task - Professional Medicine | 44.85 | Accuracy (%) | 45.7 |
| MMLU-by-task - Moral Scenarios | 24.69 | Accuracy (%) | 38.9 |
| MMLU-by-task - Security Studies | 39.59 | Accuracy (%) | 37.5 |
| MMLU-by-task - High School Chemistry | 29.06 | Accuracy (%) | 36.1 |
| MMLU-by-task - College Physics | 22.55 | Accuracy (%) | 34.3 |
| MMLU-by-task - High School Geography | 35.86 | Accuracy (%) | 33.8 |
| MMLU-by-task - Electrical Engineering | 29.66 | Accuracy (%) | 31.4 |
| MMLU-by-task - International Law | 37.19 | Accuracy (%) | 30.6 |
Interactive version: theaggregate.ai/model?slug=opt-350m · How It Works · Data refreshed daily, snapshot 2026-11-08.