GLM-5.3 Flash

Zhipu

Rank #5 of 11 · 30 PRs · 37/95 golden bugs found

40.0

F1 · +2.5 vs avg

Recall
39.0%+3.7 vs avg
Precision
41.0%-3.4 vs avg
Cost / PR
$0.080
Cost / bug found
$0.07

Confidence

95% bootstrap interval on recall (2000 resamples over the 30 PRs):29.1–49.4points. 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 35Medium 37High 19Critical 4

Category mix

Only bug/performance/security are consistent across models — the rest is free text, bucketed as other.

Bug82Performance6Security7Other0

By repository

RepoRecallPrecisionGoldensPRs
cal.com52.2%45.8%236
Discourse35.0%29.4%206
Sentry26.3%54.2%196
Keycloak23.5%24.0%176
Grafana56.3%76.2%166

Per-PR breakdown (30)