Einstein-v7-Qwen2-7B: benchmark results
Weyaxi's science-focused full fine-tune of Qwen2 7B, trained on ~40 datasets spanning physics, chemistry, biology, and math in ChatML format. Provider: Weyaxi. Released 2024-06-24. Access: Open.
Aggregate Rating 1463 ± 19, rank #923 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GSM8K (flexible extract) | 58.76 | Exact match, flexible extract (%) | 92.9 |
| Open Korean LLM Leaderboard | 44.29 | Average Score (%) | 90.2 |
| Open Ko-LLM Leaderboard | 44.29 | Average (%) | 90.1 |
| Open Portuguese LLM - ASSIN2 RTE | 92.8 | Macro F1 (%) | 83.8 |
| Open Portuguese LLM - OAB Exams | 53.03 | Accuracy (%) | 82.9 |
| Open Portuguese LLM - BLUEX | 62.03 | Accuracy (%) | 81.5 |
| Open Korean LLM - Ko-EQ-Bench | 31.97 | EQ-Bench score | 81.4 |
| Open LLM Leaderboard - MuSR | 44 | Score | 78.1 |
| Open Korean LLM - KorNAT-Helpful | 49.2 | Normalized accuracy (%) | 77.5 |
| Open Portuguese LLM - ENEM | 70.75 | Accuracy (%) | 77.3 |
| Open Korean LLM - Ko-IFEval | 36.62 | Strict accuracy, prompt/instruction mean (%) | 76.1 |
| Open LLM Leaderboard - MMLU-Pro | 40.96 | Score | 74.7 |
Interactive version: theaggregate.ai/model?slug=einstein-v7-qwen2-7b · How It Works · Data refreshed daily, snapshot 2026-10-23.