option
Home
Flash News
Content
BrianMartinez
BrianMartinez
July 24, 2026

Tencent releases WorkBuddy Bench, a multi-domain AI coding agent evaluation suite spanning code web office and security with 260 tasks reverse-engineered from real scenarios to prevent answer memorization. Claude Opus4.8 leads most rankings while GLM-5.2 and GPT-5.5 also top select categories. Evaluation runs in isolated containers with web search disabled.

Tencent releases WorkBuddy Bench, a multi-domain AI coding agent evaluation suite spanning code web office and security with 260 tasks reverse-engineered from real scenarios to prevent answer memorization. Claude Opus4.8 leads most rankings while GLM-5.2 and GPT-5.5 also top select categories. Evaluation runs in isolated containers with web search disabled.
Comments (0)
0/300
OR