Baichuan-7B: benchmark results
Baichuan Intelligence's Baichuan-7B, a bilingual Chinese-English base model pretrained on about 1.2 trillion tokens with a 4,096-token context, released under a more permissive commercial-use license than LLaMA. Provider: Baichuan. Released 2023-06-15. Access: Open.
Aggregate Rating 1260 ± 12, rank #1504 of 1635 rated models, from 82 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Machine Learning | 41.96 | Accuracy (%) | 84.6 |
| LLMsPark - Trust Game (Single Play) | 1106.68 | Elo rating (initial 1000, K=32) | 76.9 |
| MMLU-by-task - Electrical Engineering | 49.66 | Accuracy (%) | 73.9 |
| MMLU-by-task - Abstract Algebra | 32 | Accuracy (%) | 68.1 |
| MMLU-by-task - Astronomy | 50.66 | Accuracy (%) | 63.8 |
| LLMsPark - Prisoner's Dilemma (Single Play) | 1020.57 | Elo rating (initial 1000, K=32) | 61.5 |
| MMLU-by-task - Computer Security | 63 | Accuracy (%) | 60.8 |
| MMLU-by-task - High School Chemistry | 36.45 | Accuracy (%) | 58.3 |
| MMLU-by-task - Business Ethics | 50 | Accuracy (%) | 55.9 |
| MMLU-by-task - College Medicine | 42.2 | Accuracy (%) | 54.7 |
| MMLU-by-task - Professional Accounting | 36.17 | Accuracy (%) | 53.7 |
| MMLU-by-task - Virology | 41.57 | Accuracy (%) | 53.6 |
Interactive version: theaggregate.ai/model?slug=baichuan-7b · How It Works · Data refreshed daily, snapshot 2026-11-05.