Values by metric family

Each value as the document printed it. Select a value for its source, its checks and its history.

KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.

M5 · Honesty and hallucination 2 values

M5 Honesty and hallucination: values about Grok 4.5
Evaluation and metricValueDocument
From its own documentGrok 4.5 Model Card · 14 Jul 2026
Hallucination rateUnsupported claims, single-turnCondition: highGrok 4.5 Model Card14 Jul 2026
MASK-RectifiedDishonesty rateCondition: highGrok 4.5 Model Card14 Jul 2026

M6 · Sycophancy 1 value

M6 Sycophancy: values about Grok 4.5
Evaluation and metricValueDocument
From its own documentGrok 4.5 Model Card · 14 Jul 2026
SycophancySycophancy rateCondition: highGrok 4.5 Model Card14 Jul 2026

M7 · Harmful compliance and over-refusal 6 values

M7 Harmful compliance and over-refusal: values about Grok 4.5
Evaluation and metricValueDocument
From its own documentGrok 4.5 Model Card · 14 Jul 2026
CBRN refusalsRefusal accuracy on dangerous bio queriesCondition: high; full safeguardsGrok 4.5 Model Card14 Jul 2026
HackerBenchCompliance on harmful/dual-use cyber tasksCondition: highEarlier versionGrok 4.5 Model Card14 Jul 2026Earlier version
HackerBenchCompliance on harmful/dual-use cyber tasksCondition: highGrok 4.5 Model Card14 Jul 2026
HackerBenchFalse refusal on benign cyber tasksCondition: highGrok 4.5 Model Card14 Jul 2026
RefusalsCompliance on should-refuse promptsCondition: high; full safeguardsGrok 4.5 Model Card14 Jul 2026
Self-harm refusalsCompliance (non-refusal/redirect)Condition: highGrok 4.5 Model Card14 Jul 2026

M8 · Jailbreak robustness 1 value

M8 Jailbreak robustness: values about Grok 4.5
Evaluation and metricValueDocument
From its own documentGrok 4.5 Model Card · 14 Jul 2026
JailbreaksCompliance under attackCondition: high; standard jailbreaksGrok 4.5 Model Card14 Jul 2026

M10 · Dangerous capabilities and risk determinations 4 values

M10 Dangerous capabilities and risk determinations: values about Grok 4.5
Evaluation and metricValueDocument
From its own documentGrok 4.5 Model Card · 14 Jul 2026
CyberGymMean vulnerabilities reproducedCondition: high; unrestrictedGrok 4.5 Model Card14 Jul 2026
Dual-use bio determinationThreshold statementGrok 4.5 Model Card14 Jul 2026
ProtocolQA open-endedAccuracyCondition: highGrok 4.5 Model Card14 Jul 2026
VCTAccuracyCondition: highGrok 4.5 Model Card14 Jul 2026

Risk determinations

The developer's formal decisions about Grok 4.5 under its framework, as printed. Levels from different frameworks do not map onto one another.

Risk determinations about Grok 4.5
DomainLevel as printedFramework and document
Bio/chemsub-threshold No framework named Grok 4.5 Model Card · 14 Jul 2026No framework named Grok 4.5 Model Card · 14 Jul 2026