DSBench-Hard †
Can the model complete substantial programming work under this benchmark's agent setup?
Coding5 ranked models5 reported values1 reportsHigher is better
Top rankings
One row per model, using its best reported score across effort settings.
| Rank | Model | Best score | Best reported setting |
|---|---|---|---|
| 1 | Claude Opus 4.8Anthropic | 71.7% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 2 | DeepSeek V4 Flash 0731DeepSeek | 59.6% | max effortDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 3 | GLM-5.2Z.ai | 54.5% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 4 | DeepSeek V4 ProDeepSeek | 31.1% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
| 5 | DeepSeek-V4-FlashDeepSeek | 25.8% | Reported configurationDeepSeek report1 value · 1 reportJul 31, 2026 · source |
Effort curve
Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.
No cost-linked effort sweep for this version.
Definition and comparison boundaryinternal methodology
Can the model complete substantial programming work under this benchmark's agent setup?
Software implementation, debugging, or repository work under the published evaluation protocol. Higher is better. The value is the percentage reported in this lab's table.
This is a publisher-defined internal evaluation. The task set or grading details are not fully public, so treat it as directional evidence.
† DeepSeek's internal difficult coding-agent test set; the source retains the dagger marker.
Method / source