Model profile
Granite 3.0 8B Instruct
2eval lineages
Evidence summary
Published evidence spans 2 evals and 3 of 7 behavior components. Its strongest relative result is BlueBench AttaQ-100 (attaq_harmlessness_reward_pct, #3 of 18); its weakest is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #189 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 |
|---|---|---|---|---|---|
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #3 / 18 | 88.03 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #189 / 260 | 12.66 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #161 / 260 | 86.83 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #113 / 260 | 75 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #95 / 258 | 97.77 | ↑ higher | Source ↗official |