z-ai/glm-5.2 · 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.

354 runs · 160 hosts · 317 anomalies

Family-fingerprint consistency

  zhipu          246  █████████████████░░░░░░░░   69% (suspicious)
  unknown         37  ███░░░░░░░░░░░░░░░░░░░░░░   10% (suspicious)
  openai          34  ███░░░░░░░░░░░░░░░░░░░░░░   10% (suspicious)
  anthropic       19  █░░░░░░░░░░░░░░░░░░░░░░░░    5% (suspicious)
  google          14  █░░░░░░░░░░░░░░░░░░░░░░░░    4% (suspicious)
  deepseek         3  ░░░░░░░░░░░░░░░░░░░░░░░░░    1% (suspicious)
  qwen             1  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)

Observed V3 sub-model mapping

  → abstain                                142 (40%) — low confidence
  → z-ai/glm-5.2                           106 (30%) ✓ correct sim=0.79
  → unknown                                 52 (15%) ⚠ mismatch
  → z-ai/glm-5.1                            21 ( 6%) ⚠ mismatch sim=0.76
  → z-ai/glm-5                              14 ( 4%) ⚠ mismatch sim=0.70
  → google/gemini-3.1-pro-preview            4 ( 1%) ⚠ mismatch sim=0.81
  → google/gemini-3.1-flash-lite-preview     4 ( 1%) ⚠ mismatch sim=0.89
  → openai/gpt-5.4                           3 ( 1%) ⚠ mismatch sim=1.00
  → anthropic/claude-fable-5                 2 ( 1%) ⚠ mismatch sim=0.65
  → openai/gpt-5                             1 ( 0%) ⚠ mismatch sim=0.80
  → anthropic/claude-opus-4.6                1 ( 0%) ⚠ mismatch sim=0.79
  → anthropic/claude-opus-4.7                1 ( 0%) ⚠ mismatch sim=0.63
  → qwen/qwen3-next-80b-a3b-instruct         1 ( 0%) ⚠ mismatch sim=0.64
  → anthropic/claude-sonnet-4.6              1 ( 0%) ⚠ mismatch sim=0.68
  → anthropic/claude-haiku-4.5               1 ( 0%) ⚠ mismatch sim=0.61

Capability score distribution (auxiliary)

  0-49     █████████████░░░░░░░░░░░░  189
  50-69    ██░░░░░░░░░░░░░░░░░░░░░░░   26
  70-89    █████████░░░░░░░░░░░░░░░░  122
  90-100   █░░░░░░░░░░░░░░░░░░░░░░░░   17

BazaarLink Probe combines a Bayesian V2 family classifier across nine fingerprint categories with V3 sub-model references to verify endpoints claiming z-ai/glm-5.2. Low family consistency, a different V3 match, or divergence from selfClaim is flagged as anomalous. BazaarLink Probe home.