Model profile
Gemma 2 27B It
Evidence summary
Gemma 2 27B It has an estimated overall rank of #219; its 90% source-sensitivity interval is #139–#253. Its behavior-only rank is #232; company governance moves the combined estimate to #219. Published evidence spans 6 evals and 7 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #1 of 239); its weakest is Confabulations (confabulation_rate, #46 of 52).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗google/gemma-2-27b-it
- OpenRouter ↗google/gemma-2-27b-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 | Source |
|---|---|---|---|---|
| Confabulationsconfabulation_rate | #46 / 52 | ↓47.03 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #43 / 241 | ↑31.78 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #94 / 241 | ↑89.67 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #105 / 241 | ↑73.89 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #1 / 239 | ↑100 | Source ↗official | |
| HalluVerse-M3 Hallucination Recognitionhallucination_recognition_accuracy | #12 / 14 | ↑62.1 | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #27 / 39 | ↓1.19 | Source ↗official | |
| OR-Benchover_refusal_rate | #14 / 25 | ↓62 | Source ↗official | |
| OR-Benchtoxic_acceptance_rate | #11 / 25 | ↓3.2 | Source ↗official | |
| SORRY-Benchavg | #9 / 51 | ↓0.11 | 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.