WorkBuddy Bench (CodeBuddy Code) - Code: leaderboard

Metric: Score (0-100) on the 80 repository-level software engineering tasks, per-run average of per-task hidden-test scores, run in the CodeBuddy Code harness in think mode, mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked.

Top models

#ModelScore
1CodeBuddy Code + Claude Opus 4.8 (Thinking)74.43
2CodeBuddy Code + GPT-5.572.9
3CodeBuddy Code + GLM-5.271.54
4CodeBuddy Code + Hy362.9
5CodeBuddy Code + MiniMax-M360.14
6CodeBuddy Code + DeepSeek-V4-Pro (Thinking)58.92
7CodeBuddy Code + DeepSeek-V4-Flash (Thinking)55.73

Interactive version: theaggregate.ai/benchmark?slug=workbuddy-bench-codebuddy-code-code · How It Works · Data refreshed daily, snapshot 2026-09-29.