Datasets

200 Tasks

CircuitRepair

  • Claude Opus 5(Max reasoning)
    76.16%
  • DeepSeek v4 Flash Vision(Max reasoning)
    50.43%
  • Grok 4.5(Max reasoning)
    44.33%
1,000 Tasks

ReverseEngineering

  • Claude Opus 5(Max reasoning)
    81.60%
  • Grok 4.6(Max reasoning)
    70.70%
  • DeepSeek v4 Flash Vision(Max reasoning)
    59.10%