Benchmark
DSBench-Hard
DSBench-Hard is DeepSeek's internal test set of difficult coding-agent problems.
- Modality
- text
- Categories
- agents, code
- Openness
- unknown
- Source
- llm_stats
- Reported scores
- 2
Reported scores
llm_stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| DeepSeek-V4-Flash-0731 | DeepSeek | 0.596 | 2026-07-31 | evidence |
| DeepSeek-V4-Pro-0813 | DeepSeek | 0.672 | 2026-08-13 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.