Xwin-LM-70B-V0.1: benchmark results
Xwin-LM team's 70B alignment-focused model, built on Llama 2 and tuned with supervised fine-tuning, reward modeling and RLHF. Provider: Xwin-LM. Released 2023-09-15. Access: Open.
Aggregate Rating 1454 ± 12, rank #964 of 1635 rated models, from 74 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - US Foreign Policy | 94 | Accuracy (%) | 99.9 |
| MMLU-by-task - Computer Security | 80 | Accuracy (%) | 99.7 |
| MMLU-by-task - College Computer Science | 61 | Accuracy (%) | 99.6 |
| MMLU-by-task - Conceptual Physics | 70.64 | Accuracy (%) | 99.6 |
| MMLU-by-task - Jurisprudence | 84.26 | Accuracy (%) | 99.4 |
| MMLU-by-task - Security Studies | 81.22 | Accuracy (%) | 99.4 |
| MMLU-by-task - Anatomy | 65.19 | Accuracy (%) | 99.1 |
| MMLU-by-task - High School Geography | 88.89 | Accuracy (%) | 98.8 |
| MMLU-by-task - High School World History | 89.87 | Accuracy (%) | 98.8 |
| MMLU-by-task - High School Chemistry | 55.17 | Accuracy (%) | 98.6 |
| MMLU-by-task - High School European History | 84.85 | Accuracy (%) | 98.6 |
| MMLU-by-task - High School Government And Politics | 94.3 | Accuracy (%) | 98.6 |
Interactive version: theaggregate.ai/model?slug=xwin-lm-70b-v0-1 · How It Works · Data refreshed daily, snapshot 2026-10-15.