RecreationBench

RecreationBench - in-house long-horizon application-recreation benchmark evaluating hybrid-agent capabilities across five platforms: desktop (Ubuntu, macOS, Windows), mobile (Android), and web.

Category

General Agents

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for RecreationBench

Model Score Date Verified
Qwen3.8-27B
86.07%
18.08.2026 Verified
Qwen3.6-27B
29.80
18.08.2026 Unverified
Qwen3.7-Plus
1.99%
18.08.2026 Unverified
Qwen3.8-Flash-Next
100.00%
26.08.2026 Verified