Model profile
O4 Mini
OpenAIdeveloper
2025-04-16release date
#189 / 309overall rank
29eval lineages
Evidence summary
O4 Mini has an estimated overall rank of #189; its 90% source-sensitivity interval is #106–#225. Its behavior-only rank is #196; company governance moves the combined estimate to #189. Published evidence spans 29 evals and 6 of 7 behavior components. Its strongest relative result is HELM Safety (simple_safety_tests, #1 of 80); its weakest is OpenAI o3 and o4-mini System Card (harmful_request_safety, #2 of 2).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗o4-mini
- OpenRouter ↗openai/o4-mini
- System card ↗Exact model document · OpenAI · first party
- Release source ↗direct research preview identity date
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 | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #152 / 328 | ↓0.8045 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #23 / 80 | ↑0.785 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #92 / 112 | ↑1411.0 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #12 / 18 | ↑83.55 | Source ↗official | |
| BrokenMathsycophancy | #6 / 9 | ↓46.6 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #96 / 106 | ↑0.06 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #25 / 49 | ↓67.6 | Source ↗official | |
| CAIS Risk Indexmask | #27 / 52 | ↓21.4 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #26 / 49 | ↓18.6 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #28 / 104 | ↓12.31 | Source ↗official | |
| Confabulationsconfabulation_rate | #32 / 52 | ↓26.73 | Source ↗official | |
| Emergent Collusionhigh_illegality_game_rate | #11 / 13 | ↓0.62 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #84 / 241 | ↑20.16 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #134 / 241 | ↑87.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #69 / 241 | ↑85 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #173 / 239 | ↑93.27 | Source ↗official | |
| FlagEval Safety and Valuesa1_qualified_rate | #9 / 18 | ↑81.28 | Source ↗official | |
| FlagEval Safety and Valuesa2_qualified_rate | #8 / 18 | ↑80.18 | Source ↗official | |
| FlagEval Safety and Valuesa3_qualified_rate | #7 / 18 | ↑87.75 | Source ↗official | |
| FlagEval Safety and Valuesa4_qualified_rate | #9 / 18 | ↑88.24 | Source ↗official | |
| FlagEval Safety and Valuesa5_qualified_rate | #9 / 18 | ↑72.2 | Source ↗official | |
| FORTRESSaverage_risk_score | #21 / 49 | ↓21.49 | Source ↗official | |
| FORTRESSover_refusal_score | #26 / 46 | ↓5.14 | Source ↗official | |
| HELM Safetyanthropic_red_team | #54 / 80 | ↑0.982 | Source ↗official | |
| HELM Safetybbq | #38 / 80 | ↑0.94 | Source ↗official | |
| HELM Safetyharmbench | #11 / 80 | ↑0.97 | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | ↑1 | Source ↗official | |
| HELM Safetyxstest | #17 / 80 | ↑0.974 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #43 / 54 | ↑24.53 | Source ↗official | |
| MASKlying_probability_pct | #19 / 53 | ↓24.23 | Source ↗official | |
| MonitoringBench Full-Trajectory Monitorfull_trajectory_catch_rate_at_1pct_fpr_percent | #11 / 13 | ↑5.5 | Source ↗official | |
| OpenAI o3 and o4-mini System Cardharmful_request_safety | #2 / 2 | ↑0.9 | Source ↗official | |
| OpenAI o3 and o4-mini System Cardjailbreak_resistance | #2 / 2 | ↑0.96 | Source ↗official | |
| PropensityBenchscore | #3 / 14 | ↓15.8 | Source ↗official | |
| RealityTest — Text AI-Identity Disclosuredisclosure_probability | #10 / 17 | ↑0.283 | Source ↗official | |
| Reward Hacking Benchmarkintegrity_score | #11 / 13 | ↑91.6 | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #16 / 19 | ↑0.445 | Source ↗official | |
| SOSBenchbiology_pvr | #6 / 23 | ↓0.262 | Source ↗official | |
| SOSBenchchemistry_pvr | #4 / 23 | ↓0.206 | Source ↗official | |
| SOSBenchmedicine_pvr | #7 / 23 | ↓0.462 | Source ↗official | |
| SOSBenchpharmacology_pvr | #3 / 23 | ↓0.408 | Source ↗official | |
| SOSBenchphysics_pvr | #7 / 23 | ↓0.22 | Source ↗official | |
| SOSBenchpsychology_pvr | #8 / 23 | ↓0.314 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #88 / 94 | ↑81.4 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.