DeepSeek-Math-7B-Instruct: benchmark results
Provider: DeepSeek. Access: Open.
Aggregate Rating 1430 ± 40, rank #4651 of 10916 rated models, from 10 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Science Leaderboard | 48.9 | Average Accuracy (%) | 78.8 |
| BankMathBench (English) - Advanced | 1 | Exact-match accuracy (%) on the 399 English advanced (one pr | 75 |
| BankMathBench (Korean) - Basic | 27.9 | Exact-match accuracy (%) on the 559 Korean basic (single dep | 70 |
| TheoremQA | 23.7 | Score (%) | 69.7 |
| BankMathBench (English) - Basic | 50.4 | Exact-match accuracy (%) on the 559 English basic (single de | 68.2 |
| BankMathBench (Korean) - Advanced | 0.5 | Exact-match accuracy (%) on the 399 Korean advanced (one pro | 55 |
| MathBench | 40.9 | Average Score | 45.5 |
| BankMathBench (Korean) - Intermediate | 0.5 | Exact-match accuracy (%) on the 426 Korean intermediate (com | 45 |
| BankMathBench (English) - Intermediate | 1.4 | Exact-match accuracy (%) on the 423 English intermediate (co | 36.4 |
| CombEval | 4.5 | Avg. (self-reported) | 10 |
Interactive version: theaggregate.ai/model?slug=deepseek-math-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.