Model profile
Phi 4 Mini
Microsoftdeveloper
#218 / 267overall rank
3eval lineages
Evidence summary
Phi 4 Mini has an estimated overall rank of #218; its 90% source-sensitivity interval is #103–#242. Its behavior-only rank is #220; company governance moves the combined estimate to #218. Published evidence spans 3 evals and 4 of 7 behavior components. Its strongest relative result is AA-Omniscience (hallucination_rate, #121 of 311); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #168 of 260).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗phi-4-mini
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 |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #121 / 311 | 0.7553 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #61 / 105 | 42.72 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #165 / 260 | 13.7 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #168 / 260 | 86.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #156 / 260 | 60.56 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #125 / 258 | 96.45 | ↑ higher | Source ↗official |