Model Avg Tokens/Task Avg Time/Task (sec) Pass Rate (ARC+CRV)
Re_Call ACI Buffer 3,187 1.2s 100%
GPT-4 Baseline 65,200 12.4s 78–88%
Claude Baseline 59,750 10.8s 84–90%
Grok 3 Baseline 79,420 14.3s 70–82%
Token Use: 20x–25x more efficient.
Time Efficiency: 8x–12x faster.
Pass Rate: 100% accuracy.
