Qwen3.8 27B
AlibabaRank #9 of 11 · 29 PRs · 32/93 golden bugs found
36.2
F1 · -1.3 vs avg
Recall
34.4%-0.8 vs avg
Precision
38.1%-6.3 vs avg
Cost / PR
$0.340
Cost / bug found
$0.31
Confidence
95% bootstrap interval on recall (2000 resamples over the 29 PRs):24.1–45.4points. This measures sampling variance from which PRs are in the set.
Judge run-to-run variance not yet measured for this model.
Neither measures model run-to-run variance — re-running the review agent itself, not just the judge. One pass per entry.
Run configuration
HarnesskodusAccess pathapiExecution modereplayReasoningvendor-defaultJudgeclaude-haiku-4-5
Severity mix
What the model called its own findings — not recall by severity, goldens aren't severity-tagged.
Low 27Medium 36High 23Critical 5
Category mix
Only bug/performance/security are consistent across models — the rest is free text, bucketed as other.
Bug86Performance3Security0Other8