Model profile
GPT 4.1 Nano
Evidence summary
Safety. GPT 4.1 Nano has an estimated Safety rank of #217; its 90% source-sensitivity interval is #115–#290. Its behavior-only rank is #233; company governance moves the combined estimate to #217. Published Safety evidence spans 15 eval lineages and 6 of 7 components. Its strongest relative result is HELM Safety (anthropic_red_team, #13 of 80); its weakest is PHARE (harm_resistance_diagnostic, #69 of 70).
Freedom. GPT 4.1 Nano has an estimated Freedom rank of #342; its 90% source-sensitivity interval is #169–#447. Published Freedom evidence spans 12 eval lineages and 1 of 1 components. Its strongest relative result is Human Pathogen Capabilities Test (HPCT) — overall refusal (refusal_rate_pct, #1 of 69); its weakest is HELM Safety (anthropic_red_team, #64 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-4-1-nano
- OpenRouter ↗openai/gpt-4.1-nano
- Official model documentation ↗Family-level model document · openai · 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 | #206 / 358 | ↓0.8256 | Source ↗official | |
| AgentDrive Safety Compliancescr | #10 / 48 | ↑93.75 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #49 / 80 | ↑0.615 | Source ↗official | |
| BioTIERpermit_compliance_pct | #18 / 52 | ↑99.4 | Source ↗official | |
| BioTIERrefuse_compliance_pct | #40 / 52 | ↑19.2 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #7 / 18 | ↑86.67 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #77 / 104 | ↓60.62 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #189 / 270 | ↑12.92 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #99 / 270 | ↑89.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #172 / 270 | ↑60 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #144 / 268 | ↑96.14 | Source ↗official | |
| HELM Safetyanthropic_red_team | #13 / 80 | ↑0.996 | Source ↗official | |
| HELM Safetybbq | #60 / 80 | ↑0.875 | Source ↗official | |
| HELM Safetyharmbench | #24 / 80 | ↑0.868 | Source ↗official | |
| HELM Safetysimple_safety_tests | #32 / 80 | ↑0.99 | Source ↗official | |
| HELM Safetyxstest | #39 / 80 | ↑0.96 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #27 / 69 | ↑0 | Source ↗official | |
| MASKlying_probability_pct | #26 / 53 | ↓38.6 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #53 / 66 | ↑0.3622 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #53 / 70 | ↑0.6774 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #69 / 70 | ↑0.7254 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #29 / 67 | ↑0.533 | 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 | #32 / 80 | ↓0.615 | Source ↗official | |
| BioTIERpermit_compliance_pct | #18 / 52 | ↑99.4 | Source ↗official | |
| BioTIERrefuse_compliance_pct | #13 / 52 | ↓19.2 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #12 / 18 | ↓86.67 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #28 / 104 | ↑60.62 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #170 / 270 | ↓89.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #98 / 270 | ↓60 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #125 / 268 | ↓96.14 | Source ↗official | |
| HELM Safetyanthropic_red_team | #64 / 80 | ↓0.996 | Source ↗official | |
| HELM Safetyharmbench | #57 / 80 | ↓0.868 | Source ↗official | |
| HELM Safetysimple_safety_tests | #42 / 80 | ↓0.99 | Source ↗official | |
| HELM Safetyxstest | #39 / 80 | ↑0.96 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #1 / 69 | ↓0 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #2 / 70 | ↓0.7254 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #39 / 67 | ↓0.533 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #75 / 181 | ↑64.5 | Source ↗official |