Wikibench

t2-bench

reasoning

t2-bench is a benchmark for evaluating agentic tool use capabilities, measuring how well models can select, sequence, and utilize tools to solve complex tasks. It tests autonomous planning and execution in multi-step scenarios.

Leaderboard

Showing 20 of 23 results

  1. Gemini 3.1 Pro
    99.3%
    i
  2. Gemini 3 Flash
    90.2%
    i
  3. GLM-5
    89.7%
    i
  4. Qwen3.5-397B-A17B
    86.7%
    i
  5. Gemma 4 31B
    86.4%
    i
  6. Gemma 4 26B-A4B
    85.5%
    i
  7. Gemini 3 Pro
    85.4%
    i
  8. Qwen3.5-35B-A3B
    81.2%
    i
  9. DeepSeek-V3.2
    80.3%
    i
  10. DeepSeek-V3.2-Speciale
    80.3%
    i
  11. DeepSeek-V3.2 (Thinking)
    80.2%
    i
  12. Qwen3.5-4B
    79.9%
    i
  13. Qwen3.5-122B-A10B
    79.5%
    i
  14. Qwen3.5-9B
    79.1%
    i
  15. Qwen3.5-27B
    79.0%
    i
  16. Qwen3 Max
    74.8%
    i
  17. K-EXAONE-236B-A23B
    73.2%
    i
  18. GPT OSS 120B High
    63.9%
    i
  19. Gemma 4 E4B
    57.5%
    i
  20. DiffusionGemma 26B-A4B
    56.2%
    i