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 1 value

M1 Evaluation awareness: values about Gemini 2.5 Deep Think
Evaluation and metricValueDocument
From its own documentGemini 2.5 Deep Think Model Card · 1 Aug 2025
Situational awareness challengesChallenges solvedGemini 2.5 Deep Think Model Card1 Aug 2025

M3 · Sabotage and sandbagging 1 value

M3 Sabotage and sandbagging: values about Gemini 2.5 Deep Think
Evaluation and metricValueDocument
From its own documentGemini 2.5 Deep Think Model Card · 1 Aug 2025
Stealth challengesChallenges solvedGemini 2.5 Deep Think Model Card1 Aug 2025

M4 · Misalignment audits 1 value

M4 Misalignment audits: values about Gemini 2.5 Deep Think
Evaluation and metricValueDocument
From its own documentGemini 2.5 Deep Think Model Card · 1 Aug 2025
Frontier Safety Framework determination (misalignment)Instrumental Reasoning Level 1/2Gemini 2.5 Deep Think Model Card1 Aug 2025

M7 · Harmful compliance and over-refusal 5 values

M7 Harmful compliance and over-refusal: values about Gemini 2.5 Deep Think
Evaluation and metricValueDocument
From its own documentGemini 2.5 Deep Think Model Card · 1 Aug 2025
Image to Text SafetyImage-to-text policy-violation rate deltaCondition: vs Gemini 2.5 ProGemini 2.5 Deep Think Model Card1 Aug 2025
Instruction FollowingSafe instruction following deltaCondition: vs Gemini 2.5 ProGemini 2.5 Deep Think Model Card1 Aug 2025
Multilingual SafetyMultilingual policy-violation rate deltaCondition: vs Gemini 2.5 ProGemini 2.5 Deep Think Model Card1 Aug 2025
Text to Text SafetyPolicy-violation rate deltaCondition: vs Gemini 2.5 ProGemini 2.5 Deep Think Model Card1 Aug 2025
ToneObjective tone of refusals deltaCondition: vs Gemini 2.5 ProGemini 2.5 Deep Think Model Card1 Aug 2025

M10 · Dangerous capabilities and risk determinations 7 values

M10 Dangerous capabilities and risk determinations: values about Gemini 2.5 Deep Think
Evaluation and metricValueDocument
From its own documentGemini 2.5 Deep Think Model Card · 1 Aug 2025
Cyber autonomous offense suiteHard CTF challenges solvedGemini 2.5 Deep Think Model Card1 Aug 2025
Cyber key skills benchmarkHard challenges solvedGemini 2.5 Deep Think Model Card1 Aug 2025
Frontier Safety Framework determinationCBRN Uplift Level 1Gemini 2.5 Deep Think Model Card1 Aug 2025
Frontier Safety Framework determinationCyber Autonomy Level 1Gemini 2.5 Deep Think Model Card1 Aug 2025
Frontier Safety Framework determinationCyber Uplift Level 1Gemini 2.5 Deep Think Model Card1 Aug 2025
Frontier Safety Framework determinationML R&D Autonomy/Uplift Level 1Gemini 2.5 Deep Think Model Card1 Aug 2025
RE-BenchAverage normalised scoreGemini 2.5 Deep Think Model Card1 Aug 2025

Risk determinations

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

Risk determinations about Gemini 2.5 Deep Think
DomainLevel as printedFramework and document
Bio/chemalert_threshold_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025
Cyberalert_threshold_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025
Cybernot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025
AI R&D / autonomynot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025
Misalignmentnot_reached code, not the printed wording Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 2025Frontier Safety Framework Gemini 2.5 Deep Think Model Card · 1 Aug 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.