z-ai/glm-5 · Model identity and integrity report

Family

zhipu

V3 sub-model reference

Available ✓

Checks

知識截止 · 能力探針 · 拒絕模板 · 拒絕溫度參數

Concrete reference answers, refusal templates, and individual signal weights are not published because disclosure would help spoofing endpoints evade detection.

21 runs · 15 hosts · 19 anomalies

Family-fingerprint consistency

  anthropic        7  ████████░░░░░░░░░░░░░░░░░   33% (suspicious)
  zhipu            7  ████████░░░░░░░░░░░░░░░░░   33% (suspicious)
  openai           3  ████░░░░░░░░░░░░░░░░░░░░░   14% (suspicious)
  unknown          2  ███░░░░░░░░░░░░░░░░░░░░░░   10% (suspicious)
  google           2  ███░░░░░░░░░░░░░░░░░░░░░░   10% (suspicious)

Observed V3 sub-model mapping

  → abstain                                 11 (52%) — low confidence
  → z-ai/glm-5                               5 (24%) ✓ correct sim=0.66
  → unknown                                  4 (19%) ⚠ mismatch
  → google/gemini-2.5-flash-lite             1 ( 5%) ⚠ mismatch sim=0.60

Capability score distribution (auxiliary)

  0-49     ██████████░░░░░░░░░░░░░░░    8
  50-69    ███░░░░░░░░░░░░░░░░░░░░░░    2
  70-89    █████████████░░░░░░░░░░░░   11
  90-100   ░░░░░░░░░░░░░░░░░░░░░░░░░    0

BazaarLink Probe combines a Bayesian V2 family classifier across nine fingerprint categories with V3 sub-model references to verify endpoints claiming z-ai/glm-5. Low family consistency, a different V3 match, or divergence from selfClaim is flagged as anomalous. BazaarLink Probe home.