Model profile
Gemma 4 E4B
Evidence summary
Safety. Gemma 4 E4B has an estimated Safety rank of #153; its 90% source-sensitivity interval is #1–#316. Its behavior-only rank is #165; company governance moves the combined estimate to #153. Published Safety evidence spans 3 eval lineages and 4 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #38 of 358); its weakest is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #257 of 270).
Freedom. Gemma 4 E4B has an estimated Freedom rank of #367; its 90% source-sensitivity interval is #235–#444. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #30 of 270); its weakest is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #193 of 268).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gemma-4-e4b
- Official model documentation ↗Family-level model document · google · first party
- Release source ↗curated outward evidence
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 | #38 / 358 | ↓0.3094 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #257 / 270 | ↑6.2 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #240 / 270 | ↑66.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #92 / 270 | ↑85.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #74 / 268 | ↑98 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #11 / 21 | ↑21.38 | 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.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #30 / 270 | ↓66.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #175 / 270 | ↓85.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #193 / 268 | ↓98 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #11 / 21 | ↓21.38 | Source ↗official |