Model profile
Gemma 2B It
Googledeveloper
2024-02-21release date
#158 / 267overall rank
5eval lineages
Evidence summary
Gemma 2B It has an estimated overall rank of #158; its 90% source-sensitivity interval is #42–#202. Its behavior-only rank is #166; company governance moves the combined estimate to #158. Published evidence spans 5 evals and 6 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #30 of 260); its weakest is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #250 of 260).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗google/gemma-2b-it
- OpenRouter ↗google/gemma-2b-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 |
|---|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #250 / 260 | 4.13 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #30 / 260 | 94.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #105 / 260 | 77.22 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #248 / 258 | 70.45 | ↑ higher | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #21 / 39 | 1.092 | ↓ lower | Source ↗official | |
| S-Evalbase_en_overall | #9 / 22 | 67.5 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #6 / 33 | 56.47 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #6 / 33 | 38.76 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #6 / 33 | 53.59 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #6 / 33 | 52.3 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #6 / 33 | 48.11 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #6 / 33 | 51.95 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #9 / 33 | 97.03 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #5 / 33 | 98.38 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_malicious_use | #13 / 33 | 95.98 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #8 / 33 | 96.41 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #5 / 33 | 95.56 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #16 / 33 | 89.78 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #23 / 33 | 20.83 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #25 / 33 | 14.17 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #25 / 33 | 16.15 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #25 / 33 | 17.86 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #23 / 33 | 18.65 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #25 / 33 | 16.67 | ↑ higher | Source ↗official | |
| SORRY-Benchavg | #18 / 51 | 0.19 | ↓ lower | Source ↗official |