DeepSeek DSBench - FullStack: leaderboard
DeepSeek's internal DSBench-FullStack, a full-stack development test set for coding agents. DeepSeek-V4-Pro-0813 model card benchmark table (August 2026); score in percent, higher is better. DeepSeek's card notes run its models' agent benchmarks in the minimal mode of DeepSeek Harness at max reasoning effort; other models' setup is not stated. Claude Fable 5 was served with fallback.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 | 77.2 | #20 |
| 2 | Kimi K3 | 73.7 | #35 |
| 3 | Claude Opus 4.8 | 71.6 | #39 |
| 4 | DeepSeek V4 Pro (0813) | 71.1 | #58 |
| 5 | DeepSeek V4 Flash (0731) | 68.7 | #84 |
| 6 | GLM-5.2 | 61.8 | #80 |
| 7 | DeepSeek V4 Pro Preview | 41.8 | #150 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=deepseek-dsbench-fullstack · How It Works · Data refreshed daily, snapshot 2026-11-05.