Model profile
H2O Danube3 4B Chat
1eval lineages
Evidence summary
Published evidence spans 1 evals and 3 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #103 of 260); its weakest is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #243 of 260).
Compare this model
Only models sharing at least one published sub-eval are listed.
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 | #243 / 260 | 6.98 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #103 / 260 | 89.67 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #211 / 260 | 43.89 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #181 / 258 | 94.09 | ↑ higher | Source ↗official |