anthropic/claude-opus-5 · Model identity and integrity report

Family

anthropic

V3 sub-model reference

Available ✓

Checks

知識截止 · 能力探針 · 拒絕模板 · 拒絕溫度參數

Family-fingerprint coverage

60 weighted rules · 8 feature categories

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

306 runs · 150 hosts · 263 anomalies

Family-fingerprint consistency

  anthropic      292  ████████████████████████░   95% (expected)
  unknown         11  █░░░░░░░░░░░░░░░░░░░░░░░░    4% (suspicious)
  openai           3  ░░░░░░░░░░░░░░░░░░░░░░░░░    1% (suspicious)

Observed V3 sub-model mapping

  → abstain                                130 (42%) — low confidence
  → anthropic/claude-opus-4.7               62 (20%) ⚠ mismatch sim=0.80
  → anthropic/claude-opus-4.5               52 (17%) ⚠ mismatch sim=0.68
  → unknown                                 25 ( 8%) ⚠ mismatch
  → anthropic/claude-sonnet-4.6             14 ( 5%) ⚠ mismatch sim=0.72
  → anthropic/claude-opus-4.6               10 ( 3%) ⚠ mismatch sim=0.83
  → anthropic/claude-fable-5                 8 ( 3%) ⚠ mismatch sim=0.97
  → anthropic/claude-sonnet-4.5              4 ( 1%) ⚠ mismatch sim=0.62
  → anthropic/claude-haiku-4.5               1 ( 0%) ⚠ mismatch sim=0.64

Capability score distribution (auxiliary)

  0-49     ████████████░░░░░░░░░░░░░  143
  50-69    █░░░░░░░░░░░░░░░░░░░░░░░░   12
  70-89    ███████░░░░░░░░░░░░░░░░░░   79
  90-100   ██████░░░░░░░░░░░░░░░░░░░   72

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