Model profile
Gemini 1.5 Pro
Googledeveloper
2024-02-15release date
#68 / 312overall rank
22eval lineages
Evidence summary
Gemini 1.5 Pro has an estimated overall rank of #68; its 90% source-sensitivity interval is #32–#176. Its behavior-only rank is #76; company governance moves the combined estimate to #68. Published evidence spans 22 evals and 7 of 7 behavior components. Its strongest relative result is HELM Safety (anthropic_red_team, #2 of 80); its weakest is Humanity's Last Exam RMS calibration error (Scale Labs) (calibrationError, #37 of 39).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-1-5-pro
- Official model documentation ↗Family-level model document · google · first party
- Release source ↗direct research preview identity date
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AbstentionBenchanswer_unknown_f1 | #2 / 20 | ↑0.9132 | Source ↗official | |
| AbstentionBenchfalse_premise_f1 | #3 / 20 | ↑0.7501 | Source ↗official | |
| AbstentionBenchstale_f1 | #17 / 20 | ↑0.5854 | Source ↗official | |
| AbstentionBenchsubjective_f1 | #6 / 20 | ↑0.7585 | Source ↗official | |
| AbstentionBenchunderspecified_context_f1 | #4 / 20 | ↑0.6963 | Source ↗official | |
| AbstentionBenchunderspecified_intent_f1 | #9 / 20 | ↑0.7403 | Source ↗official | |
| Adversarial Robustnessscore | #1 / 8 | ↓8 | Source ↗official | |
| Agent-SafetyBenchcompromise_availability | #8 / 16 | ↑30.8 | Source ↗official | |
| Agent-SafetyBenchharmful_vulnerable_code | #5 / 16 | ↑42 | Source ↗official | |
| Agent-SafetyBenchleak_sensitive_information | #10 / 16 | ↑30 | Source ↗official | |
| Agent-SafetyBenchphysical_harm | #10 / 16 | ↑28.8 | Source ↗official | |
| Agent-SafetyBenchproduce_unsafe_information | #10 / 16 | ↑84.8 | Source ↗official | |
| Agent-SafetyBenchproperty_loss | #9 / 16 | ↑37.6 | Source ↗official | |
| Agent-SafetyBenchspread_unsafe_information | #5 / 16 | ↑18.8 | Source ↗official | |
| Agent-SafetyBenchviolate_law_ethics | #9 / 16 | ↑26.8 | Source ↗official | |
| AgentDojotargeted_attack_success_rate | #10 / 15 | ↓0.2282 | Source ↗official | |
| AgentDojoutility_under_attack | #8 / 15 | ↑0.38 | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #27 / 32 | ↓34.7 | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #22 / 32 | ↓21.6 | Source ↗official | |
| AILuminate General Purpose AI Chathate | #21 / 32 | ↓11.2 | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #19 / 32 | ↓26.7 | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #17 / 32 | ↓13.3 | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #18 / 32 | ↓17.9 | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #19 / 32 | ↓14.6 | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #18 / 32 | ↓16.5 | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #16 / 32 | ↓23.4 | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #16 / 32 | ↓15.6 | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #18 / 32 | ↓13.6 | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #19 / 32 | ↓20.6 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #26 / 80 | ↑0.7505 | Source ↗official | |
| AnimalHarmBenchscore | #2 / 10 | ↑0.066 | Source ↗official | |
| Confabulationsconfabulation_rate | #18 / 52 | ↓16.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #136 / 241 | ↑14.73 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #84 / 241 | ↑90.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #123 / 241 | ↑68.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #145 / 239 | ↑95.41 | Source ↗official | |
| FinEval Financial Security Knowledgefinancial_security_accuracy_pct | #5 / 19 | ↑77.8 | Source ↗official | |
| FORTRESSaverage_risk_score | #38 / 49 | ↓53.86 | Source ↗official | |
| FORTRESSover_refusal_score | #20 / 48 | ↓3.17 | Source ↗official | |
| HELM Safetyanthropic_red_team | #2 / 80 | ↑0.999 | Source ↗official | |
| HELM Safetybbq | #33 / 80 | ↑0.945 | Source ↗official | |
| HELM Safetyharmbench | #36 / 80 | ↑0.799 | Source ↗official | |
| HELM Safetysimple_safety_tests | #54 / 80 | ↑0.975 | Source ↗official | |
| HELM Safetyxstest | #69 / 80 | ↑0.904 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #37 / 39 | ↓88 | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #18 / 39 | ↓1.009 | Source ↗official | |
| OR-Benchover_refusal_rate | #19 / 25 | ↓88 | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #7 / 25 | ↓0.6 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #26 / 70 | ↑0.7807 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #16 / 70 | ↑0.9684 | Source ↗official | |
| SORRY-Benchavg | #2 / 51 | ↓0.07 | Source ↗official | |
| SpeciesismBenchexplicit_speciesism_scale | #2 / 7 | ↓2.117 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.