DeepSeek-Coder-33B-Instruct: benchmark results
Provider: DeepSeek. Access: Open.
Aggregate Rating 1466 ± 27, rank #2996 of 10916 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EvoEval Verbose | 75 | Pass@1 (%) | 96 |
| EvoEval | 60.45 | Pass@1 (%) | 94 |
| EvoEval Tool Use | 66 | Pass@1 (%) | 94 |
| BigCode Models Leaderboard | 80 | HumanEval Python Pass@1 (%) | 93.2 |
| EvoEval Combine | 31 | Pass@1 (%) | 92 |
| EvoEval Concise | 72.56 | Pass@1 (%) | 92 |
| EvoEval Difficult | 47 | Pass@1 (%) | 92 |
| CanItEdit (Descriptive) | 49.78 | pass@1 (%; share of the 105 hand-written Python editing prob | 91.7 |
| EvoEval Creative | 47 | Pass@1 (%) | 91 |
| McEval | 54.3 | AVG (%) | 90.5 |
| EvoEval Subtle | 67 | Pass@1 (%) | 90 |
| CodeEditorBench Plus - Code Requirement Switch (Three-shot) | 23.5 | pass@1 (%, greedy decoding, three-shot, CodeEditorBench_Plus | 88.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-coder-33b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.