stack/eval/results/mbpp.json

{
  "model": "Stack 2.9 (estimate)",
  "benchmark": "MBPP",
  "pass@1": 0.8,
  "pass@10": 0.85,
  "pass@100": 0.88,
  "note": "Estimate based on Qwen2.5-Coder-32B. Actual after training.",
  "source": "Qwen2.5-Coder technical report"
}