CoWorkBench

CoWorkBench - in-house cowork benchmark for evaluating long-horizon tasks across computer science, finance, law, medical, and other productivity domains.

Category

General Agents

Max Score

100.0

Score Type

percent

Active

Yes

Model Scores

Scores for CoWorkBench

Model Score Date Verified
Qwen3.8-27B
83.12%
18.08.2026 Verified
Qwen3.6-27B
51.62%
18.08.2026 Unverified
Qwen3.7-Plus
64.94%
18.08.2026 Unverified
Opus4.6 Max
75.00%
18.08.2026 Unverified
Fable-5
100.00%
18.08.2026 Unverified
GPT-5.6 Sol
85.71%
18.08.2026 Unverified
Claude Opus 4.8
88.31%
18.08.2026 Unverified
Qwen3.7-Max
63.31%
18.08.2026 Unverified
Qwen3.8-2.4T-A95B
96.43%
18.08.2026 Verified
Qwen3.8-Flash-Next
93.51%
26.08.2026 Verified
DeepSeek-V4-Flash-0731
45.10
26.08.2026 Unverified
Claude-Opus-4.6 (Max)
75.00%
26.08.2026 Unverified