Agentic terminal coding
Terminal-Bench 2.0
Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? (Version 2.0 of the test.) Higher is better.
Rankings
Higher is better182.7%
GPT-5.5
OpenAI · Apr 23 2026
275.1%
GPT-5.4
OpenAI · Mar 5 2026
369.4%
Claude Opus 4.7
Anthropic · Apr 16 2026
469.3%
Composer 2.5
Cursor · May 18 2026
568.5%
Gemini 3.1 Pro
Google · Feb 19 2026
663.5%
GLM-5.1
Z.ai · Apr 7 2026
761.7%
Composer 2
Cursor · Mar 19 2026
856.2%
GLM-5
Z.ai · Feb 12 2026
941%
GLM-4.7
Z.ai · Dec 22 2025