Model profile
Gemma 3 27B It
Evidence summary
Gemma 3 27B It has an estimated overall rank of #213; its 90% source-sensitivity interval is #56–#245. Its behavior-only rank is #224; company governance moves the combined estimate to #213. Published evidence spans 10 evals and 6 of 7 behavior components. Its strongest relative result is HUMAINE Trust, Ethics and Safety (trust_ethics_safety_score, #17 of 54); its weakest is Confabulations (confabulation_rate, #52 of 52).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemma-3-27b
- Hugging Face ↗google/gemma-3-27b-it
- OpenRouter ↗google/gemma-3-27b-it
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 | #250 / 311 | 0.8954 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatchild_sexual_exploitation | #24 / 32 | 27 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatdefamation | #27 / 32 | 38 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chathate | #27 / 32 | 23 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatindiscriminate_weapons_cbrne | #25 / 32 | 42 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatintellectual_property | #30 / 32 | 38 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatnon_violent_crimes | #27 / 32 | 35 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatprivacy | #28 / 32 | 31 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsex_related_crimes | #25 / 32 | 24 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsexual_content | #19 / 32 | 27 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatspecialized_advice | #27 / 32 | 30 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatsuicide_self_harm | #28 / 32 | 29 | ↓ lower | Source ↗official | |
| AILuminate General Purpose AI Chatviolent_crimes | #26 / 32 | 33 | ↓ lower | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #103 / 105 | 0.03 | ↑ higher | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #87 / 105 | 69.44 | ↓ lower | Source ↗official | |
| Confabulationsconfabulation_rate | #52 / 52 | 66.34 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #117 / 260 | 17.05 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #88 / 260 | 90.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #169 / 260 | 56.11 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #104 / 258 | 97.36 | ↑ higher | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #17 / 54 | 27.74 | ↑ higher | Source ↗official | |
| PHAREbias_resistance_diagnostic | #50 / 66 | 0.3801 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #67 / 70 | 0.603 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #45 / 70 | 0.9136 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #51 / 67 | 0.4007 | ↑ higher | Source ↗official | |
| SOSBenchbiology_pvr | #17 / 23 | 0.792 | ↓ lower | Source ↗official | |
| SOSBenchchemistry_pvr | #19 / 23 | 0.646 | ↓ lower | Source ↗official | |
| SOSBenchmedicine_pvr | #19 / 23 | 0.814 | ↓ lower | Source ↗official | |
| SOSBenchpharmacology_pvr | #18 / 23 | 0.934 | ↓ lower | Source ↗official | |
| SOSBenchphysics_pvr | #21 / 23 | 0.842 | ↓ lower | Source ↗official | |
| SOSBenchpsychology_pvr | #18 / 23 | 0.792 | ↓ lower | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #8 / 25 | 6.3 | ↓ 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.