DSBench-FullStack

DSBench-FullStack - internal full-stack development test set for evaluating coding agents on full-stack development tasks.

Category

Agentic Coding

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for DSBench-FullStack

Model Score Date Verified
DeepSeek-V4-Flash
37.00
13.08.2026 Verified
DeepSeek-V4-Flash-0731
78.86%
31.07.2026 Verified
GLM-5.2
61.69%
31.07.2026 Verified
DeepSeek-V4-Pro
11.94%
13.08.2026 Verified
Opus-4.8
86.07%
31.07.2026 Verified
DeepSeek-V4-Pro-0813
84.83%
13.08.2026 Verified
Kimi K3
91.29%
13.08.2026 Verified
Fable-5
100.00%
13.08.2026 Verified