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 under 128k tokens, returning nodes reachable at specified depths.

Leaderboard

Showing 11 of 11 results

  1. GPT-5.2
    94.0%
    i
  2. GPT-5.4
    93.0%
    i
  3. GPT-5
    78.3%
    i
  4. GPT-5.4 mini
    76.3%
    i
  5. GPT-5.4 nano
    73.4%
    i
  6. GPT-4.5
    72.3%
    i
  7. GPT-4.1
    61.7%
    i
  8. GPT-4.1 mini
    61.7%
    i
  9. o3-mini
    51.0%
    i
  10. GPT-4o
    41.7%
    i
  11. GPT-4.1 nano
    25.0%
    i