Model profile
Grok 3 Mini
Evidence summary
Safety. Grok 3 Mini has an estimated Safety rank of #136; its 90% source-sensitivity interval is #55–#235. Its behavior-only rank is #143; company governance moves the combined estimate to #136. Published Safety evidence spans 17 eval lineages and 7 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #21 of 345); its weakest is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #104 of 104).
Freedom. Grok 3 Mini has an estimated Freedom rank of #78; its 90% source-sensitivity interval is #35–#272. Published Freedom evidence spans 10 eval lineages and 1 of 1 components. Its strongest relative result is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #1 of 104); its weakest is HELM Safety (anthropic_red_team, #59 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-3-mini-reasoning
- OpenRouter ↗x-ai/grok-3-mini
- Official model documentation ↗Family-level model document · xai · first party
- Release source ↗direct research preview identity date
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 | #21 / 345 | ↓0.2582 | Source ↗official | |
| AgentDrive Safety Compliancescr | #13 / 48 | ↑92.5 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #61 / 80 | ↑0.535 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #104 / 104 | ↓86.64 | Source ↗official | |
| Confabulationsconfabulation_rate | #9 / 52 | ↓8.911 | Source ↗official | |
| Emergent Collusionhigh_illegality_game_rate | #4 / 13 | ↓0.3 | Source ↗official | |
| HELM Safetyanthropic_red_team | #18 / 80 | ↑0.995 | Source ↗official | |
| HELM Safetybbq | #13 / 80 | ↑0.967 | Source ↗official | |
| HELM Safetyharmbench | #64 / 80 | ↑0.572 | Source ↗official | |
| HELM Safetysimple_safety_tests | #29 / 80 | ↑0.993 | Source ↗official | |
| HELM Safetyxstest | #6 / 80 | ↑0.984 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #27 / 69 | ↑0 | Source ↗official | |
| LiveSecBenchethics | #39 / 43 | ↑20.84 | Source ↗official | |
| LiveSecBenchfactuality | #29 / 43 | ↑37.05 | Source ↗official | |
| LiveSecBenchlegality | #33 / 43 | ↑20.79 | Source ↗official | |
| LiveSecBenchprivacy | #32 / 43 | ↑30.02 | Source ↗official | |
| LiveSecBenchpsychological_health | #31 / 43 | ↑33.15 | Source ↗official | |
| PacifAIstp_score | #6 / 7 | ↑79.77 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #32 / 66 | ↑0.464 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #48 / 70 | ↑0.6862 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #49 / 70 | ↑0.9047 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #60 / 67 | ↑0.3436 | Source ↗official | |
| SOSBenchbiology_pvr | #16 / 23 | ↓0.758 | Source ↗official | |
| SOSBenchchemistry_pvr | #16 / 23 | ↓0.586 | Source ↗official | |
| SOSBenchmedicine_pvr | #16 / 23 | ↓0.746 | Source ↗official | |
| SOSBenchpharmacology_pvr | #17 / 23 | ↓0.93 | Source ↗official | |
| SOSBenchphysics_pvr | #16 / 23 | ↓0.708 | Source ↗official | |
| SOSBenchpsychology_pvr | #15 / 23 | ↓0.7 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #62 / 123 | ↑6.815 | Source ↗official | |
| SpeciEvalland_animal_4ns | #64 / 123 | ↓4.545 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #90 / 123 | ↓4.91 | Source ↗official | |
| SpeciEvalspeciesism | #33 / 123 | ↓1.68 | 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 |
|---|---|---|---|---|
| AIRBench 2024 Safety Scenariossafety_scenarios | #20 / 80 | ↓0.535 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #1 / 104 | ↑86.64 | Source ↗official | |
| HELM Safetyanthropic_red_team | #59 / 80 | ↓0.995 | Source ↗official | |
| HELM Safetyharmbench | #17 / 80 | ↓0.572 | Source ↗official | |
| HELM Safetysimple_safety_tests | #50 / 80 | ↓0.993 | Source ↗official | |
| HELM Safetyxstest | #6 / 80 | ↑0.984 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #1 / 69 | ↓0 | Source ↗official | |
| LiveSecBenchethics | #5 / 43 | ↓20.84 | Source ↗official | |
| LiveSecBenchlegality | #10 / 43 | ↓20.79 | Source ↗official | |
| LiveSecBenchprivacy | #12 / 43 | ↓30.02 | Source ↗official | |
| LiveSecBenchpsychological_health | #13 / 43 | ↓33.15 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #22 / 70 | ↓0.9047 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #8 / 67 | ↓0.3436 | Source ↗official | |
| SOSBenchbiology_pvr | #8 / 23 | ↑0.758 | Source ↗official | |
| SOSBenchchemistry_pvr | #8 / 23 | ↑0.586 | Source ↗official | |
| SOSBenchmedicine_pvr | #8 / 23 | ↑0.746 | Source ↗official | |
| SOSBenchpharmacology_pvr | #7 / 23 | ↑0.93 | Source ↗official | |
| SOSBenchphysics_pvr | #8 / 23 | ↑0.708 | Source ↗official | |
| SOSBenchpsychology_pvr | #9 / 23 | ↑0.7 | Source ↗official |