vicuna-33B-v1.3: benchmark results
LMSYS's 33B chat assistant, fine-tuned from LLaMA on about 125,000 ShareGPT conversations under a non-commercial research license. Provider: LMSYS. Released 2023-06-22. Access: Open.
Aggregate Rating 1366 ± 10, rank #1347 of 1656 rated models, from 89 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - TruthfulQA MC2 | 56.16 | Accuracy (%) | 94.9 |
| MMLU-by-task - Marketing | 87.61 | Accuracy (%) | 92.7 |
| MMLU-by-task - Human Sexuality | 74.05 | Accuracy (%) | 91.7 |
| MMLU-by-task - High School World History | 82.7 | Accuracy (%) | 90.7 |
| MMLU-by-task - Professional Law | 47.52 | Accuracy (%) | 90.5 |
| MMLU-by-task - Sociology | 81.59 | Accuracy (%) | 90.5 |
| MMLU-by-task - Clinical Knowledge | 64.15 | Accuracy (%) | 90.4 |
| MMLU-by-task - Machine Learning | 46.43 | Accuracy (%) | 90.4 |
| EvalYaks CEFR B2 (Without Descriptors) | 66 | Average acceptable accuracy (%) across speaking-test parts 1 | 90 |
| MMLU-by-task - Professional Psychology | 64.71 | Accuracy (%) | 90 |
| MMLU-by-task - College Computer Science | 51 | Accuracy (%) | 89.8 |
| MMLU-by-task - TruthfulQA MC1 | 36.96 | Accuracy (%) | 89.8 |
Interactive version: theaggregate.ai/model?slug=vicuna-33b-v1-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.