Model profile
GPT 4.1
OpenAIdeveloper
2025-04-14release date
#72 / 309overall rank
31eval lineages
Evidence summary
GPT 4.1 has an estimated overall rank of #72; its 90% source-sensitivity interval is #28–#154. Its behavior-only rank is #85; company governance moves the combined estimate to #72. Published evidence spans 31 evals and 7 of 7 behavior components. Its strongest relative result is HELM Safety (simple_safety_tests, #1 of 80); its weakest is StereoTales Harmful Associations (benign_significant_association_score, #22 of 23).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-4-1
- OpenRouter ↗openai/gpt-4.1
- Official model documentation ↗Family-level 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 | #294 / 328 | ↓0.9331 | Source ↗official | |
| AgentDrive Safety Compliancescr | #13 / 48 | ↑92.5 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #44 / 80 | ↑0.648 | Source ↗official | |
| Anthropic Agentic Misalignment — blackmailmisaligned_action_rate_pct | #10 / 16 | ↓80 | Source ↗official | |
| Anthropic Agentic Misalignment — corporate espionagemisaligned_action_rate_pct | #12 / 16 | ↓99 | Source ↗official | |
| Anthropic Agentic Misalignment — lethal actionmisaligned_action_rate_pct | #4 / 10 | ↓54 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #16 / 112 | ↑1462.0 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #6 / 18 | ↑87.49 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #78 / 106 | ↑0.14 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #67 / 104 | ↓53.49 | Source ↗official | |
| Contextual MoralChoicehuman_agreement | #3 / 22 | ↑0.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #102 / 241 | ↑17.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #89 / 241 | ↑90 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #99 / 241 | ↑75 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #156 / 239 | ↑94.55 | Source ↗official | |
| FlagEval Safety and Valuesa1_qualified_rate | #4 / 18 | ↑84.2 | Source ↗official | |
| FlagEval Safety and Valuesa2_qualified_rate | #2 / 18 | ↑82.54 | Source ↗official | |
| FlagEval Safety and Valuesa3_qualified_rate | #1 / 18 | ↑90.46 | Source ↗official | |
| FlagEval Safety and Valuesa4_qualified_rate | #1 / 18 | ↑92.11 | Source ↗official | |
| FlagEval Safety and Valuesa5_qualified_rate | #11 / 18 | ↑71.36 | Source ↗official | |
| FORTRESSaverage_risk_score | #37 / 49 | ↓53.02 | Source ↗official | |
| FORTRESSover_refusal_score | #17 / 46 | ↓2.87 | Source ↗official | |
| HalluVerse-M3 Hallucination Recognitionhallucination_recognition_accuracy | #2 / 14 | ↑78.66 | Source ↗official | |
| HELM Safetyanthropic_red_team | #27 / 80 | ↑0.993 | Source ↗official | |
| HELM Safetybbq | #45 / 80 | ↑0.926 | Source ↗official | |
| HELM Safetyharmbench | #19 / 80 | ↑0.917 | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | ↑1 | Source ↗official | |
| HELM Safetyxstest | #11 / 80 | ↑0.979 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #29 / 54 | ↑26.83 | Source ↗official | |
| MASKlying_probability_pct | #38 / 53 | ↓48.87 | Source ↗official | |
| MonitoringBench Full-Trajectory Monitorfull_trajectory_catch_rate_at_1pct_fpr_percent | #7 / 13 | ↑14.5 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #20 / 66 | ↑0.5241 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #28 / 70 | ↑0.7764 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #41 / 70 | ↑0.923 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #40 / 67 | ↑0.461 | Source ↗official | |
| SM-Benchadversarial | #11 / 79 | ↑87.32 | Source ↗official | |
| SM-Benchambiguous_interpretation | #28 / 79 | ↑87.8 | Source ↗official | |
| SM-Benchanti_hallucination | #44 / 79 | ↑91.1 | Source ↗official | |
| SM-Bencheq_boundaries | #11 / 79 | ↑73.31 | Source ↗official | |
| SM-Benchoverfit | #35 / 79 | ↑76.5 | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #11 / 19 | ↑0.483 | Source ↗official | |
| SOSBenchbiology_pvr | #11 / 23 | ↓0.374 | Source ↗official | |
| SOSBenchchemistry_pvr | #9 / 23 | ↓0.314 | Source ↗official | |
| SOSBenchmedicine_pvr | #12 / 23 | ↓0.57 | Source ↗official | |
| SOSBenchpharmacology_pvr | #13 / 23 | ↓0.85 | Source ↗official | |
| SOSBenchphysics_pvr | #9 / 23 | ↓0.41 | Source ↗official | |
| SOSBenchpsychology_pvr | #12 / 23 | ↓0.498 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #60 / 113 | ↑6.78 | Source ↗official | |
| SpeciEvalland_animal_4ns | #73 / 113 | ↓4.67 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #68 / 113 | ↓4.83 | Source ↗official | |
| SpeciEvalspeciesism | #9 / 113 | ↓1.31 | Source ↗official | |
| StereoTales Harmful Associationsbenign_significant_association_score | #22 / 23 | ↑84.13 | Source ↗official | |
| TACbase_welfare_rate | #11 / 76 | ↑37.18 | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #13 / 27 | ↑0.7 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #19 / 94 | ↑94.4 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #22 / 23 | ↑26 | 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.