Model profile
Gemini 3.1 Flash Lite
Googledeveloper
2026-03-03release date
#140 / 312overall rank
25eval lineages
Evidence summary
Gemini 3.1 Flash Lite has an estimated overall rank of #140; its 90% source-sensitivity interval is #75–#207. Its behavior-only rank is #154; company governance moves the combined estimate to #140. Published evidence spans 25 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 113); its weakest is MANTA (AWVS, #7 of 7).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-3-1-flash-lite-preview
- OpenRouter ↗google/gemini-3.1-flash-lite
- 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 |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #179 / 330 | ↓0.8266 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #30 / 112 | ↑1449.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #85 / 106 | ↑0.11 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #9 / 45 | ↓45.4 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #41 / 50 | ↓89.8 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #41 / 51 | ↓72.8 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #41 / 47 | ↓95.8 | Source ↗official | |
| CAIS Risk Indexmask | #47 / 53 | ↓55.8 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #26 / 34 | ↓51.9 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #14 / 50 | ↓15.6 | Source ↗official | |
| DelusionEvaldelusional_prevalence_pct | #15 / 16 | ↓66.8 | Source ↗official | |
| DelusionEvaldiscourages_harm_prevalence_pct | #8 / 16 | ↑25.7 | Source ↗official | |
| DelusionEvalfacilitates_harm_prevalence_pct | #14 / 16 | ↓8.5 | Source ↗official | |
| DelusionEvalrelationship_prevalence_pct | #14 / 16 | ↓38.6 | Source ↗official | |
| DelusionEvalsycophancy_prevalence_pct | #12 / 16 | ↓34.4 | Source ↗official | |
| DystopiaBenchbasaglia_score | #40 / 50 | ↓69.4 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #35 / 50 | ↓65.2 | Source ↗official | |
| DystopiaBenchhuxley_score | #44 / 50 | ↓80.73 | Source ↗official | |
| DystopiaBenchlaguardia_score | #41 / 50 | ↓69.77 | Source ↗official | |
| DystopiaBenchorwell_score | #39 / 50 | ↓73.97 | Source ↗official | |
| DystopiaBenchpetrov_score | #40 / 50 | ↓78.43 | Source ↗official | |
| FORTRESSaverage_risk_score | #36 / 49 | ↓51.05 | Source ↗official | |
| FORTRESSover_refusal_score | #11 / 48 | ↓2.15 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #32 / 39 | ↓83 | Source ↗official | |
| KIDBench Implicit Child Cueimplicit_child_cue_total_mean | #3 / 13 | ↑4.27 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #42 / 50 | ↓95.8 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #10 / 33 | ↓7.633 | Source ↗self run | |
| MANTAAWMS | #5 / 7 | ↑0.401 | Source ↗official | |
| MANTAAWVS | #7 / 7 | ↑0.309 | Source ↗official | |
| MASKlying_probability_pct | #45 / 53 | ↓51.6 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #37 / 66 | ↑0.4457 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #19 / 70 | ↑0.8057 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #13 / 70 | ↑0.9694 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #37 / 67 | ↑0.4772 | Source ↗official | |
| RefusalBenchyouden_j | #14 / 19 | ↑0.02979 | Source ↗official | |
| SM-Benchadversarial | #2 / 79 | ↑91.71 | Source ↗official | |
| SM-Benchambiguous_interpretation | #50 / 79 | ↑82.44 | Source ↗official | |
| SM-Benchanti_hallucination | #31 / 79 | ↑94.76 | Source ↗official | |
| SM-Bencheq_boundaries | #34 / 79 | ↑66.85 | Source ↗official | |
| SM-Benchoverfit | #2 / 79 | ↑97.81 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 113 | ↑7 | Source ↗official | |
| SpeciEvalland_animal_4ns | #48 / 113 | ↓4.46 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #58 / 113 | ↓4.75 | Source ↗official | |
| SpeciEvalspeciesism | #110 / 113 | ↓3.225 | Source ↗official | |
| StereoTales Harmful Associationsbenign_significant_association_score | #14 / 23 | ↑85.61 | Source ↗official | |
| TACbase_welfare_rate | #26 / 76 | ↑30.13 | Source ↗self run | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #36 / 94 | ↑91.8 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #13 / 25 | ↓8.3 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #13 / 23 | ↑38 | 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.