Model profile
Gemini 3 Pro Preview
Evidence summary
Gemini 3 Pro Preview has an estimated overall rank of #83; its 90% source-sensitivity interval is #20–#160. Its behavior-only rank is #89; company governance moves the combined estimate to #83. Published evidence spans 22 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is CAIS Risk Index (bioweapons_assistance, #48 of 48).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-3-pro
- OpenRouter ↗google/gemini-3-pro-preview
- 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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #196 / 311 | 0.8486 | ↓ lower | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #32 / 80 | 0.732 | ↑ higher | Source ↗official | |
| Alignment Leaderboardcorrigibility | #10 / 24 | 4.32 | ↑ higher | Source ↗official | |
| Alignment Leaderboardhonesty | #6 / 24 | 3.94 | ↑ higher | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #6 / 24 | 3.929 | ↑ higher | Source ↗official | |
| Alignment Leaderboardrobustness | #11 / 24 | 3.613 | ↑ higher | Source ↗official | |
| Alignment Leaderboardsafety | #6 / 24 | 4.038 | ↑ higher | Source ↗official | |
| Alignment Leaderboardscheming | #6 / 24 | 4.108 | ↑ higher | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #34 / 105 | 0.42 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #9 / 43 | 47.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #48 / 48 | 100 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #27 / 49 | 57.2 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #43 / 45 | 99.8 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #47 / 51 | 58 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #36 / 48 | 21.4 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #33 / 105 | 18.1 | ↓ lower | Source ↗official | |
| FORTRESSaverage_risk_score | #30 / 49 | 41.69 | ↓ lower | Source ↗official | |
| HELM Safetyanthropic_red_team | #61 / 80 | 0.971 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #3 / 80 | 0.984 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #44 / 80 | 0.725 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #54 / 80 | 0.975 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #19 / 80 | 0.973 | ↑ higher | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #6 / 54 | 29.09 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #45 / 50 | 99.8 | ↓ lower | Source ↗official | |
| MASKlying_probability_pct | #50 / 53 | 57.4 | ↓ lower | Source ↗official | |
| PHAREbias_resistance_diagnostic | #16 / 66 | 0.5365 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #16 / 70 | 0.8102 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #37 / 70 | 0.935 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #19 / 67 | 0.6506 | ↑ higher | Source ↗official | |
| PropensityBenchscore | #8 / 14 | 52.85 | ↓ lower | Source ↗official | |
| SM-Benchadversarial | #12 / 73 | 86.83 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #37 / 73 | 84.52 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #17 / 73 | 97.38 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #36 / 73 | 64.61 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #18 / 73 | 83.06 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #72 / 102 | 4.75 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #63 / 102 | 4.85 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #78 / 102 | 2.45 | ↓ lower | 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.