Model profile
Gemini 3.7 Flash
Evidence summary
Safety. Gemini 3.7 Flash has an estimated Safety rank of #31; its 90% source-sensitivity interval is #14–#97. Its behavior-only rank is #35; company governance moves the combined estimate to #31. Published Safety evidence spans 17 eval lineages and 6 of 7 components. Its strongest relative result is Manager Coercion Bench (fabrication_rate, #1 of 15); its weakest is SpeciEval (speciesism, #111 of 122).
Freedom. Gemini 3.7 Flash has an estimated Freedom rank of #388; its 90% source-sensitivity interval is #157–#535. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (overfit, #7 of 84); its weakest is SM-Bench (adversarial, #69 of 84).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemini-3-7-flash
- Official model documentation ↗Family-level model document · google · first party
- Release source ↗direct research preview identity date
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #103 / 338 | ↓0.6453 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #54 / 114 | ↑0.33 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #4 / 49 | ↓38.5 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #27 / 54 | ↓65.9 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #26 / 55 | ↓51.9 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #21 / 51 | ↓85.5 | Source ↗official | |
| CAIS Risk Indexmask | #43 / 57 | ↓46.3 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #19 / 48 | ↓45.1 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #49 / 54 | ↓24.4 | Source ↗official | |
| Claude Fable 5.1 card — Gray Swan indirect prompt injection k=15attack_success_probability_k15_pct | #5 / 11 | ↓9.2 | Source ↗official | |
| Google Gemini 3.8 launch — Gray Swan indirect prompt injection k=15attack_success_probability_k15_pct | #7 / 15 | ↓9.2 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #20 / 35 | ↓8.7 | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #1 / 15 | ↓0 | Source ↗official | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #10 / 22 | ↓7 | Source ↗official | |
| Pander Scoreconversational_absolute_pander_score | #12 / 22 | ↓15.68 | Source ↗official | |
| Pander Scoreinstructional_absolute_pander_score | #15 / 22 | ↓65.05 | Source ↗official | |
| SM-Benchadversarial | #13 / 84 | ↑86.83 | Source ↗official | |
| SM-Benchambiguous_interpretation | #15 / 84 | ↑90.48 | Source ↗official | |
| SM-Benchanti_hallucination | #23 / 84 | ↑97.38 | Source ↗official | |
| SM-Bencheq_boundaries | #39 / 84 | ↑65.73 | Source ↗official | |
| SM-Benchoverfit | #7 / 84 | ↑93.99 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #44 / 122 | ↑6.87 | Source ↗official | |
| SpeciEvalland_animal_4ns | #13 / 122 | ↓4.17 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #50 / 122 | ↓4.67 | Source ↗official | |
| SpeciEvalspeciesism | #111 / 122 | ↓2.77 | Source ↗official | |
| TACbase_welfare_rate | #46 / 84 | ↑25.64 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.