Zephyr-7B-β: benchmark results
Provider: HuggingFace. Access: Open.
Aggregate Rating 1312 ± 23, rank #8496 of 10916 rated models, from 73 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenCompass Language - Content Summarization - English (CompassBench 2404) | 30.3 | Score (%) | 70 |
| OpenCompass Code - MBPP - Chinese (CompassBench 2404) | 30.2 | Score (%) | 46.7 |
| OpenCompass Code - MBPP - English (CompassBench 2404) | 38.9 | Score (%) | 40 |
| OpenCompass LLM - Knowledge - English (CompassBench 2404) | 66.8 | Score (%) | 36.7 |
| OpenCompass Subjective - Creation (CompassBench 2404) | 17.9 | Score (%) | 36 |
| OpenCompass Math - High - English (CompassBench 2404) | 9.6 | Score (%) | 34 |
| OpenCompass Agent - CIBench - Chinese (CompassBench 2404) | 23.6 | Score (%) | 33.3 |
| OpenCompass LLM - Agent - Chinese (CompassBench 2404) | 37 | Score (%) | 33.3 |
| OpenCompass Reasoning - Inductive - English (CompassBench 2404) | 21 | Score (%) | 33.3 |
| OpenCompass Agent - T-Eval - Chinese (CompassBench 2404) | 50.5 | Score (%) | 30 |
| OpenCompass Language - Intention Recognition - Chinese (CompassBench 2404) | 10.7 | Score (%) | 30 |
| OpenCompass Reasoning - Abductive - Chinese (CompassBench 2404) | 29.6 | Score (%) | 30 |
Interactive version: theaggregate.ai/model?slug=zephyr-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.