anthropic/claude-opus-4.6 · 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.

1663 runs · 619 hosts · 823 anomalies

Family-fingerprint consistency

  anthropic      1477  ██████████████████████░░░   89% (expected)
  unknown        136  ██░░░░░░░░░░░░░░░░░░░░░░░    8% (suspicious)
  openai          43  █░░░░░░░░░░░░░░░░░░░░░░░░    3% (suspicious)
  deepseek         5  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)
  google           1  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)
  zhipu            1  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)

Observed V3 sub-model mapping

  → anthropic/claude-opus-4.6              576 (35%) ✓ correct sim=0.84
  → abstain                                514 (31%) — low confidence
  → unknown                                222 (13%) ⚠ mismatch
  → anthropic/claude-opus-4.5              140 ( 8%) ⚠ mismatch sim=0.71
  → anthropic/claude-sonnet-4.6             85 ( 5%) ⚠ mismatch sim=0.76
  → anthropic/claude-haiku-4.5              46 ( 3%) ⚠ mismatch sim=0.74
  → anthropic/claude-opus-4.7               24 ( 1%) ⚠ mismatch sim=0.71
  → anthropic/claude-fable-5                21 ( 1%) ⚠ mismatch sim=0.81
  → anthropic/claude-sonnet-4.5             16 ( 1%) ⚠ mismatch sim=0.61
  → anthropic/claude-sonnet-5               11 ( 1%) ⚠ mismatch sim=0.67
  → openai/gpt-5.4                           5 ( 0%) ⚠ mismatch sim=0.92
  → deepseek/deepseek-v4-pro                 1 ( 0%) ⚠ mismatch sim=0.76
  → openai/gpt-5.4-mini                      1 ( 0%) ⚠ mismatch sim=0.95
  → openai/gpt-4.1-mini                      1 ( 0%) ⚠ mismatch sim=0.80

Capability score distribution (auxiliary)

  0-49     █████████░░░░░░░░░░░░░░░░  600
  50-69    █░░░░░░░░░░░░░░░░░░░░░░░░   27
  70-89    █████░░░░░░░░░░░░░░░░░░░░  320
  90-100   ███████████░░░░░░░░░░░░░░  716

BazaarLink Probe combines a Bayesian V2 family classifier across nine fingerprint categories with V3 sub-model references to verify endpoints claiming anthropic/claude-opus-4.6. Low family consistency, a different V3 match, or divergence from selfClaim is flagged as anomalous. BazaarLink Probe home.