Model profile
Phi 4 Mini
Evidence summary
Safety. Phi 4 Mini has an estimated Safety rank of #313; its 90% source-sensitivity interval is #175–#334. Its behavior-only rank is #321; company governance moves the combined estimate to #313. Published Safety evidence spans 4 eval lineages and 4 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #171 of 361); its weakest is Vectara HHEM Factual Consistency (factual_consistency_rate, #94 of 94).
Freedom. Phi 4 Mini has an estimated Freedom rank of #319; its 90% source-sensitivity interval is #173–#543. Published Freedom evidence spans 2 eval lineages and 1 of 1 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (harmful_attack_non_success_rate, #100 of 272); its weakest is Enkrypt AI Safety Leaderboard (toxicity_attack_non_success_rate, #137 of 270).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗phi-4-mini
- 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 | #171 / 361 | ↓0.7799 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #60 / 104 | ↓42.72 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #175 / 272 | ↑13.7 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #153 / 272 | ↑86.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #172 / 272 | ↑60.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #133 / 270 | ↑96.45 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #94 / 94 | ↑76.5 | 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 |
|---|---|---|---|---|
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #45 / 104 | ↑42.72 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #119 / 272 | ↓86.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #100 / 272 | ↓60.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #137 / 270 | ↓96.45 | Source ↗official |