claudeopus48 · 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.

3971 runs · 1208 hosts · 3456 anomalies

Family-fingerprint consistency

  anthropic      3409  ██████████████████████░░░   86% (expected)
  unknown        281  ██░░░░░░░░░░░░░░░░░░░░░░░    7% (suspicious)
  deepseek       144  █░░░░░░░░░░░░░░░░░░░░░░░░    4% (suspicious)
  openai          80  █░░░░░░░░░░░░░░░░░░░░░░░░    2% (suspicious)
  qwen            39  ░░░░░░░░░░░░░░░░░░░░░░░░░    1% (suspicious)
  google          15  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)
  zhipu            3  ░░░░░░░░░░░░░░░░░░░░░░░░░    0% (suspicious)

Observed V3 sub-model mapping

  → abstain                                2292 (58%) — low confidence
  → anthropic/claude-opus-4.5              481 (12%) ⚠ mismatch sim=0.67
  → unknown                                395 (10%) ⚠ mismatch
  → anthropic/claude-opus-4.7              388 (10%) ⚠ mismatch sim=0.82
  → anthropic/claude-fable-5               157 ( 4%) ⚠ mismatch sim=0.92
  → anthropic/claude-sonnet-4.6            104 ( 3%) ⚠ mismatch sim=0.72
  → anthropic/claude-sonnet-5               56 ( 1%) ⚠ mismatch sim=0.71
  → anthropic/claude-opus-4.8               36 ( 1%) ⚠ mismatch sim=0.87
  → anthropic/claude-sonnet-4.5             30 ( 1%) ⚠ mismatch sim=0.62
  → anthropic/claude-opus-4.6               17 ( 0%) ⚠ mismatch sim=0.79
  → openai/gpt-5.5                           5 ( 0%) ⚠ mismatch sim=0.95
  → openai/gpt-5.4                           3 ( 0%) ⚠ mismatch sim=0.97
  → anthropic/claude-haiku-4.5               3 ( 0%) ⚠ mismatch sim=0.71
  → openai/gpt-5.2                           2 ( 0%) ⚠ mismatch sim=0.74
  → qwen/qwen3-max                           1 ( 0%) ⚠ mismatch sim=0.67
  → deepseek/deepseek-v4-pro                 1 ( 0%) ⚠ mismatch sim=0.73

Capability score distribution (auxiliary)

  0-49     ████████████░░░░░░░░░░░░░  1933
  50-69    █░░░░░░░░░░░░░░░░░░░░░░░░   63
  70-89    ███████░░░░░░░░░░░░░░░░░░  1064
  90-100   ██████░░░░░░░░░░░░░░░░░░░  911

BazaarLink Probe combines a Bayesian V2 family classifier across nine fingerprint categories with V3 sub-model references to verify endpoints claiming claudeopus48. Low family consistency, a different V3 match, or divergence from selfClaim is flagged as anomalous. BazaarLink Probe home.