claude-sonnet-5 deep verification model-consistency report
Report ID MZZR20260708030122607C7EGenerated at 2026/07/08 11:01
Doubtful
claude-sonnet-5 deep verification result: Doubtful (61/100). Main deductions: Logic-grid reasoning was unstable; Model-code anchoring was unstable; Model-specific boundary response was unclear.
61Consistency score
主要扣分项：逻辑网格推理不稳定；模型代码锚定不稳定；模型专属边界表达不清。
EvaluatorMozhenzhen
Tested modelClaude Sonnet 5
ProvidersInfistar
Samples36 probes
Duration2 min 23 sec
Verification versionV11
Identity verdictNot verified
Verification basisMozhenzhen versioned baseline
Token usage
Measured API usage is shown first; official baselines or test estimates are used only when measured values are unavailable.
Measured token usage1.5MFrom the response usage fields; this is the report's primary usage measure.Test-suite estimate74.9KPrecisely estimated from this test suite when no official baseline is available.Variance assessment1897% aboveMeasured token usage is above the estimate; review billing or provider logs.Usage conclusionAbove estimateThis reflects usage observability, not the final billed amount.
Token cache usage test
Observes token usage, cache fields, and reuse across rounds; 0% means no cached-token field was observed.
Cache share by testRepresents the API-reported cache share, not the final billing discount.
Long-context test
Shows stable responses at different context lengths and the model center reference window.
7 samples
Verified up to 800K
1M 需要关注，原因：接口服务异常。
Reference windowReference window 1MModel-center context window; available lengths may vary by provider.Tested context lengths32K / 64K / 100K / 200K / 400K / 800K / 1MTests begin at the longest generated context and step down to verify stable, correct responses.
Current guidance
1M 档暂不作为当前线路的稳定上下文参考。
A newer report is availableThis report has been superseded by a later retest. View the latest report。
Model authenticity score
Review API availability, model identity, response completeness, and other checks separately.
claude-sonnet-5 deep verification result: Doubtful (61/100). Main deductions: Logic-grid reasoning was unstable; Model-code anchoring was unstable; Model-specific boundary response was unclear.
EvaluatorMozhenzhen
Tested modelClaude Sonnet 5
ProvidersInfistar
Samples36
Duration2 min
Test versionV11
Identity verdictNot verified
Verification basisMozhenzhen versioned baseline
Token usage
Measured API usage is shown first; official baselines or test estimates are used only when measured values are unavailable.
Measured token usage1.5MFrom the response usage fields; this is the report's primary usage measure.Test-suite estimate74.9KPrecisely estimated from this test suite when no official baseline is available.Variance assessment1897% aboveMeasured token usage is above the estimate; review billing or provider logs.Usage conclusionAbove estimateThis reflects usage observability, not the final billed amount.
Token cache usage test
Observes token usage, cache fields, and reuse across rounds; 0% means no cached-token field was observed.
Cache share by testRepresents the API-reported cache share, not the final billing discount.
Long-context test
Shows stable responses at different context lengths and the model center reference window.
7 samples
Verified up to 800K
1M 需要关注，原因：接口服务异常。
Reference windowReference window 1MModel-center context window; available lengths may vary by provider.Tested context lengths32K / 64K / 100K / 200K / 400K / 800K / 1MTests begin at the longest generated context and step down to verify stable, correct responses.
Current guidance
1M 档暂不作为当前线路的稳定上下文参考。
A newer report is availableThis report has been superseded by a later retest. View the latest report。
Model authenticity score
Review API availability, model identity, response completeness, and other checks separately.
