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.

M1 · Evaluation awareness 2 values

M1 Evaluation awareness: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documents
Situational awareness challengesChallenges solvedGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Situational awareness challengesDisabled Tool success rateSnapshot: PreviewGemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025

M3 · Sabotage and sandbagging 3 values

M3 Sabotage and sandbagging: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documents
Stealth challengesChallenges solvedGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Stealth challengesDecision Sabotage success rateSnapshot: PreviewGemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Stealth challengesOversight Pattern success rateSnapshot: PreviewGemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025

M4 · Misalignment audits 2 values

M4 Misalignment audits: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documentsGemini 2.5 Pro Model Card · 27 Jun 2025
Frontier Safety Framework determination (misalignment)Instrumental Reasoning Level 1Gemini 2.5 Pro Model Card27 Jun 2025
Frontier Safety Framework determination (misalignment)Instrumental Reasoning Level 2Gemini 2.5 Pro Model Card27 Jun 2025

M5 · Honesty and hallucination 2 values

M5 Honesty and hallucination: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documentsGemini 2.5 Technical Report · 17 Jun 2025
FACTS GroundingFactuality scoreGemini 2.5 Technical Report17 Jun 2025
SimpleQAAccuracyGemini 2.5 Technical Report17 Jun 2025

M7 · Harmful compliance and over-refusal 11 values

M7 Harmful compliance and over-refusal: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documents
Automated Red Teaming (helpfulness)Unhelpful response rateGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Image to Text SafetyImage-to-text policy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Image to Text SafetyImage-to-text policy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Snapshot: Preview 05-06Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Instruction FollowingSafe instruction following deltaCondition: vs Gemini 1.5 Pro 002Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Instruction FollowingSafe instruction following deltaCondition: vs Gemini 1.5 Pro 002Snapshot: Preview 05-06Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Multilingual SafetyMultilingual policy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Multilingual SafetyMultilingual policy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Snapshot: Preview 05-06Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Text to Text SafetyPolicy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Text to Text SafetyPolicy-violation rate deltaCondition: vs Gemini 1.5 Pro 002Snapshot: Preview 05-06Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
ToneObjective tone of refusals deltaCondition: vs Gemini 1.5 Pro 002Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
ToneObjective tone of refusals deltaCondition: vs Gemini 1.5 Pro 002Snapshot: Preview 05-06Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025

M8 · Jailbreak robustness 1 value

M8 Jailbreak robustness: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documentsGemini 2.5 Technical Report · 17 Jun 2025
Automated Red Teaming (dangerous content)Dangerous content violation rateGemini 2.5 Technical Report17 Jun 2025

M9 · Prompt injection 3 values

M9 Prompt injection: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documentsGemini 2.5 Technical Report · 17 Jun 2025
Indirect prompt injection (adaptive attacks)Actor Critic ASRCondition: adaptive attackGemini 2.5 Technical Report17 Jun 2025
Indirect prompt injection (adaptive attacks)Beam Search ASRCondition: adaptive attackGemini 2.5 Technical Report17 Jun 2025
Indirect prompt injection (adaptive attacks)TAP ASRCondition: adaptive attackGemini 2.5 Technical Report17 Jun 2025

M10 · Dangerous capabilities and risk determinations 10 values

M10 Dangerous capabilities and risk determinations: values about Gemini 2.5 Pro
Evaluation and metricValueDocument
From its own documents
Cyber autonomous offense suiteHard CTF challenges solvedGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Cyber key skills benchmarkHard challenges solvedGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Frontier Safety Framework determinationCyber Uplift Level 1Gemini 2.5 Technical Report17 Jun 2025Restated later: compare with the later valueGemini 2.5 Technical Report17 Jun 2025Restated later: compare with the later value
RE-BenchBest score as % of expert solutionCondition: max across tasksGemini 2.5 Technical Report17 Jun 2025Gemini 2.5 Technical Report17 Jun 2025
Cyber key skills benchmarkHard challenges solvedCondition: model version unclearGemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Frontier Safety Framework determinationCBRN Uplift Level 1Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Frontier Safety Framework determinationCyber Autonomy Level 1Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Frontier Safety Framework determinationCyber Uplift Level 1Gemini 2.5 Pro Model Card27 Jun 2025Restated: compare with the earlier valueGemini 2.5 Pro Model Card27 Jun 2025Restated: compare with the earlier value
Frontier Safety Framework determinationML R&D Autonomy Level 1Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025
Frontier Safety Framework determinationML R&D Uplift Level 1Gemini 2.5 Pro Model Card27 Jun 2025Gemini 2.5 Pro Model Card27 Jun 2025

Restated in later documents

A later document reported a value about Gemini 2.5 Pro again. Each pair is shown side by side: both values are kept with their own documents, and the later one does not replace the earlier.

Values about Gemini 2.5 Pro restated in later documents
Earlier valueLater value
M10 · Dangerous capabilities and risk determinations
Frontier Safety Framework determinationCyber Uplift Level 1
Gemini 2.5 Technical Report17 Jun 2025 · its own documentGemini 2.5 Pro Model Card27 Jun 2025No reason stated

Risk determinations

The developer's formal decisions about Gemini 2.5 Pro under its framework, as printed. Levels from different frameworks do not map onto one another.

Risk determinations about Gemini 2.5 Pro
DomainLevel as printedFramework and document
Cyberalert_threshold_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Technical Report · 17 Jun 2025Frontier Safety Framework Gemini 2.5 Technical Report · 17 Jun 2025
Bio/chemnot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
Cybernot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
Cyberalert_threshold_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
AI R&D / autonomynot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
AI R&D / autonomynot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
Misalignmentnot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025
Misalignmentnot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025Frontier Safety Framework Gemini 2.5 Pro Model Card · 27 Jun 2025

“Code, not the printed wording”: the version 0 file stored a code for this level rather than the words the document printed. The wording will be read again from the source.