Wikibench

Graphwalks BFS >128k

reasoning

A graph reasoning benchmark that evaluates language models' ability to perform breadth-first search (BFS) operations on graphs with context length over 128k tokens, testing long-context reasoning capabilities.

Leaderboard

Showing 8 of 8 results

  1. Claude Mythos Preview
    80.0%
    i
  2. Claude Opus 4.8
    68.1%
    i
  3. Claude Opus 4.6
    61.5%
    i
  4. GPT-5.5
    45.4%
    i
  5. GPT-5.4
    21.4%
    i
  6. GPT-4.1
    19.0%
    i
  7. GPT-4.1 mini
    15.0%
    i
  8. GPT-4.1 nano
    2.9%
    i