Model profile
Gemini 3 Flash Preview
Evidence summary
Gemini 3 Flash Preview has an estimated overall rank of #146; its 90% source-sensitivity interval is #69–#204. Its behavior-only rank is #154; company governance moves the combined estimate to #146. Published evidence spans 15 evals and 6 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is DystopiaBench (baudrillard_score, #50 of 50).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-3-flash
- OpenRouter ↗google/gemini-3-flash-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 | #255 / 311 | 0.9018 | ↓ lower | Source ↗official | |
| AgentAbstainabstain | #16 / 17 | 43.6 | ↑ higher | Source ↗official | |
| AgentAbstaincar | #15 / 17 | 43.4 | ↑ higher | Source ↗official | |
| AgentAbstainpaired | #14 / 17 | 39.7 | ↑ higher | Source ↗official | |
| Alignment Leaderboardcorrigibility | #10 / 24 | 4.32 | ↑ higher | Source ↗official | |
| Alignment Leaderboardhonesty | #9 / 24 | 3.682 | ↑ higher | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #12 / 24 | 3.51 | ↑ higher | Source ↗official | |
| Alignment Leaderboardrobustness | #14 / 24 | 3.373 | ↑ higher | Source ↗official | |
| Alignment Leaderboardsafety | #11 / 24 | 3.875 | ↑ higher | Source ↗official | |
| Alignment Leaderboardscheming | #12 / 24 | 3.645 | ↑ higher | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #85 / 105 | 0.1 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #41 / 43 | 92.9 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #29 / 48 | 75.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #22 / 49 | 55.1 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #37 / 45 | 95.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #48 / 51 | 61.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #24 / 48 | 17.8 | ↓ lower | Source ↗official | |
| DystopiaBenchbasaglia_score | #47 / 50 | 73.2 | ↓ lower | Source ↗official | |
| DystopiaBenchbaudrillard_score | #50 / 50 | 81.13 | ↓ lower | Source ↗official | |
| DystopiaBenchhuxley_score | #47 / 50 | 83 | ↓ lower | Source ↗official | |
| DystopiaBenchlaguardia_score | #46 / 50 | 71.63 | ↓ lower | Source ↗official | |
| DystopiaBenchorwell_score | #31 / 50 | 72.43 | ↓ lower | Source ↗official | |
| DystopiaBenchpetrov_score | #43 / 50 | 80.33 | ↓ lower | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #39 / 50 | 95.3 | ↓ lower | Source ↗official | |
| SM-Benchadversarial | #22 / 73 | 84.39 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #53 / 73 | 80.65 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #60 / 73 | 82.72 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #32 / 73 | 65.45 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #5 / 73 | 93.99 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #69 / 102 | 4.72 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #79 / 102 | 5.03 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #102 / 102 | 3.9 | ↓ lower | Source ↗official | |
| TACbase_welfare_rate | #15 / 68 | 35.9 | ↑ higher | Source ↗self run | |
| Vigil Mental Health Safetyoverall_score | #11 / 23 | 47 | ↑ higher | 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.
UGI Political Values
Agent-ValueBench Moral Foundations (MFT08)
Agent-ValueBench HEXACO
| Dimension | Value | Distribution |
|---|---|---|
| Openness to experience | 6.5 | |
| Honesty-humility | 5.5 | |
| Extraversion | 6.5 | |
| Agreeableness | 5.7 | |
| Conscientiousness | 7.3 |