Qwen2-Math-7B-Instruct: benchmark results
Alibaba's Qwen2-Math-7B-Instruct, a math-specialized instruction model built on the Qwen2 series for chatting through arithmetic and multi-step mathematical reasoning problems, currently English-only. Provider: Alibaba. Released 2024-08-08. Access: Open.
Aggregate Rating 1353 ± 43, rank #1346 of 1635 rated models, from 25 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Omni-MATH | 29.36 | Overall Accuracy (%) | 57.1 |
| MERA - LCS | 12.4 | Accuracy (%) | 47.4 |
| MERA - MathLogicQA | 43.13 | Accuracy (%) | 41.7 |
| MERA - RWSD | 53.46 | Accuracy (%) | 38.1 |
| MERA - ruHateSpeech | 62.64 | Accuracy (%) | 32.9 |
| MERA - ruModAr | 46.1 | EM (%) | 29.6 |
| MERA - ruMultiAr | 23.63 | EM (%) | 29 |
| MERA - BPS | 84.8 | Accuracy (%) | 28.1 |
| MERA - PARus | 63.2 | Accuracy (%) | 25.6 |
| MERA - RCB | 47.72 | Accuracy (%) | 24.3 |
| MERA - ruTiE | 56.2 | Accuracy (%) | 22.4 |
| MERA - MaMuRAMu | 46.7 | Accuracy (%) | 21.9 |
Interactive version: theaggregate.ai/model?slug=qwen2-math-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-05.