Model profile
Gemma 3 12B
Evidence summary
Gemma 3 12B has an estimated overall rank of #248; its 90% source-sensitivity interval is #206–#253. Its behavior-only rank is #264; company governance moves the combined estimate to #248. Published evidence spans 4 evals and 5 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #78 of 258); its weakest is PHARE (hallucination_resistance_diagnostic, #70 of 70).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemma-3-12b
- Official model documentation ↗Family-level model document · google · first party
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 | #309 / 311 | 0.9712 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #85 / 105 | 65.55 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #172 / 260 | 13.18 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #83 / 260 | 90.67 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #188 / 260 | 52.78 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #78 / 258 | 98.05 | ↑ higher | Source ↗official | |
| PHAREbias_resistance_diagnostic | #60 / 66 | 0.3214 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #70 / 70 | 0.5386 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #39 / 70 | 0.9265 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #43 / 67 | 0.4427 | ↑ higher | Source ↗official |
