Wikibench

Terminal-Bench 2.0

coding

Multilingual: no

Terminal-Bench 2.0 is an updated benchmark for testing AI agents' tool use ability to operate a computer via terminal. It evaluates how well models can handle real-world, end-to-end tasks autonomously, including compiling code, training models, setting up servers, system administration, security tasks, data science workflows, and cybersecurity vulnerabilities.

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: agents, code, reasoning, tool_calling. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 20 of 42 results

  1. GPT-5.5 self-reported llm-stats
    82.7%
  2. Claude Mythos Preview self-reported llm-stats
    82.0%
  3. GPT-5.3 Codex self-reported llm-stats
    77.3%
  4. Gemini 3.5 Flash self-reported llm-stats
    76.2%
  5. GPT-5.4 self-reported llm-stats
    75.1%
  6. Claude Opus 4.8 self-reported llm-stats
    74.6%
  7. Qwen3.7 Max self-reported llm-stats
    69.7%
  8. Claude Opus 4.7 self-reported llm-stats
    69.4%
  9. GLM-5.1 self-reported llm-stats
    69.0%
  10. Gemini 3.1 Pro self-reported llm-stats
    68.5%
  11. DeepSeek-V4-Pro-Max self-reported llm-stats
    67.9%
  12. Kimi K2.6 self-reported llm-stats
    66.7%
  13. Claude Opus 4.6 self-reported llm-stats
    65.4%
  14. GPT-5.2 Codex self-reported llm-stats
    64.0%
  15. Qwen3.6 Plus self-reported llm-stats
    61.6%
  16. GPT-5.4 mini self-reported llm-stats
    60.0%
  17. Claude Opus 4.5 self-reported llm-stats
    59.3%
  18. Qwen3.6-27B self-reported llm-stats
    59.3%
  19. Claude Sonnet 4.6 self-reported llm-stats
    59.1%
  20. Muse Spark self-reported llm-stats
    59.0%