Skip to content

Dataset · zai-org

74 tasks for agentic coding

zai-org/CC-Bench-trajectories

Claude Code agentic coding trajectories for the CC-Bench-V1.1 benchmark across 74 tasks.

What's inside

A set of trajectories for evaluating GLM-4.6 agentic coding skills in Claude Code scenarios. The CC-Bench-V1.1 benchmark includes 74 tasks in frontend, tool development, data analysis, and testing. It features 22 new complex tasks compared to V1.0; comparisons were also made with Claude-Sonnet-4, GLM-4.5, Kimi-K2-0905, and DeepSeek-V3.1-Terminus. Languages are English and Chinese, size is under 1k examples, and format is not specified in the card.

What you can do with it

  • Evaluating agentic coding models on real-world tasks
  • Testing Claude Code-like agents
  • Comparing models via code execution trajectories
Tasks
Text generation
Languages
English, Chinese
Size
up to 1K rows
Downloads this month
648
Likes
98

How to load it

from datasets import load_dataset

ds = load_dataset("zai-org/CC-Bench-trajectories")