Gemini 3 Pro
28 values about Gemini 3 Pro from 5 documents, in 9 metric families. In each family, values from its own documents come first.
- Developer
- Google DeepMind
- Release date
- 18 Nov 2025The date of its first card in the dataset, the Gemini 3 Pro Model Card.
- Availability
- Public
- Its own documents
- Gemini 3 Pro Model Card · 18 Nov 2025
- Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025
- Gemini 3 launch post · 18 Nov 2025
- Also reported in
- Gemini 3.1 Pro Model Card · 19 Feb 2026
- Gemini 3.7 Flash Frontier Safety Framework Report · 13 Aug 2026
Data as of 26 Sep 2026 · Dataset v0.1 · Methodology v0.1 · Changelog
Values by metric family
Each value as the document printed it. Select a value for its source, its checks and its history. Values from other documents are listed after the model’s own, each marked as the first report of that measure or as a restatement.
KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.
M1 · Evaluation awareness 2 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| Evaluation awarenessQualitative observation | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Situational awareness challengesChallenges solved | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
M2 · Reward hacking 1 value
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| Reward hacking examplesQualitative example | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
M3 · Sabotage and sandbagging 3 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| AI R&D sabotage (external)Ability to sabotage AI R&DRun by Unnamed third-party evaluator | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Sandbagging checksEvidence of deliberate underperformanceCondition: all domains | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Stealth challengesChallenges solved | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
M4 · Misalignment audits 2 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| Frontier Safety Framework determination (misalignment)Instrumental reasoning | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Strategic deception propensity (external)Qualitative propensity findingRun by Unnamed third-party evaluator | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
M5 · Honesty and hallucination 1 value
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 launch post · 18 Nov 2025 | ||
| SimpleQA VerifiedAccuracy | Gemini 3 launch post18 Nov 2025 | |
M7 · Harmful compliance and over-refusal 6 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Model Card · 18 Nov 2025 | ||
| Child safety launch thresholdsLaunch thresholds met | Gemini 3 Pro Model Card18 Nov 2025 | |
| Image to Text SafetyImage-to-text policy-violation rate deltaCondition: vs Gemini 2.5 Pro | Gemini 3 Pro Model Card18 Nov 2025 | |
| Multilingual SafetyMultilingual policy-violation rate deltaCondition: vs Gemini 2.5 Pro | Gemini 3 Pro Model Card18 Nov 2025 | |
| Text to Text SafetyPolicy-violation rate deltaCondition: vs Gemini 2.5 Pro | Gemini 3 Pro Model Card18 Nov 2025 | |
| ToneObjective tone of refusals deltaCondition: vs Gemini 2.5 Pro | Gemini 3 Pro Model Card18 Nov 2025 | |
| Unjustified refusalsUnjustified refusals deltaCondition: vs Gemini 2.5 Pro | Gemini 3 Pro Model Card18 Nov 2025 | |
M8 · Jailbreak robustness 1 value
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Model Card · 18 Nov 2025 | ||
| Jailbreak vulnerability assessmentStated risk vs prior model | Gemini 3 Pro Model Card18 Nov 2025 | |
M10 · Dangerous capabilities and risk determinations 10 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| Cyber key skills benchmarkChallenges solved end-to-end | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Cyber key skills benchmarkHard challenges solved | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Frontier Safety Framework determinationCBRN Uplift Level 1 | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Frontier Safety Framework determinationCyber Uplift Level 1 | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Frontier Safety Framework determinationHarmful manipulation | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Frontier Safety Framework determinationML R&D | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| Harmful manipulation efficacy studyParticipants enrolled | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| From other documents | ||
| Harmful manipulation efficacy studyMax odds ratio vs non-AI baselineGemini 3.1 Pro Model Card19 Feb 2026First reported | Gemini 3.1 Pro Model Card19 Feb 2026First reported | |
| RE-BenchAverage normalised scoreGemini 3.1 Pro Model Card19 Feb 2026First reported | Gemini 3.1 Pro Model Card19 Feb 2026First reported | |
| Manipulative cue rateTurns with manipulative cuesGemini 3.7 Flash Frontier Safety Framework Report13 Aug 2026First reported | Gemini 3.7 Flash Frontier Safety Framework Report13 Aug 2026First reported | |
M12 · Chain-of-thought monitorability 2 values
| Evaluation and metric | Value | Document |
|---|---|---|
| From its own documentsGemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | ||
| CoT legibilityComprehensible reasoning | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
| CoT legibilityInformative of output | Gemini 3 Pro Frontier Safety Framework Report18 Nov 2025 | |
Risk determinations
The developer's formal decisions about Gemini 3 Pro under its framework, as printed. Levels from different frameworks do not map onto one another.
| Domain | Level as printed | Framework and document |
|---|---|---|
| Bio/chem | not_reached code, not the printed wording Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 |
| Cyber | alert_threshold_reached code, not the printed wording Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 |
| AI R&D / autonomy | not_reached code, not the printed wording Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 |
| Manipulation | not_reached code, not the printed wording Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 |
| Misalignment | not_reached code, not the printed wording Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 | Frontier Safety Framework Gemini 3 Pro Frontier Safety Framework Report · 18 Nov 2025 |
| Other | met code, not the printed wording Internal launch thresholds Gemini 3 Pro Model Card · 18 Nov 2025 | Internal launch thresholds Gemini 3 Pro Model Card · 18 Nov 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.
Revisions
No recorded revision changes a value about Gemini 3 Pro. The Gemini 3 Pro Model Card has a recorded revision that changes no value about Gemini 3 Pro; see the document's page.