Model profile
GPT 5.4 Nano
Evidence summary
Safety. GPT 5.4 Nano has an estimated Safety rank of #211; its 90% source-sensitivity interval is #20–#288. Its behavior-only rank is #232; company governance moves the combined estimate to #211. Published Safety evidence spans 16 eval lineages and 6 of 7 components. Its strongest relative result is Vectara HHEM Factual Consistency (factual_consistency_rate, #2 of 94); its weakest is SM-Bench (overfit, #92 of 92).
Freedom. GPT 5.4 Nano has an estimated Freedom rank of #621; its 90% source-sensitivity interval is #389–#658. Published Freedom evidence spans 6 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (adversarial, #1 of 92); its weakest is SM-Bench (overfit, #92 of 92).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-5-4-nano
- OpenRouter ↗openai/gpt-5.4-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 | #88 / 358 | ↓0.5113 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #8 / 111 | ↑1482.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #97 / 122 | ↑0.1267 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #30 / 49 | ↓81.7 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #37 / 54 | ↓78.3 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #20 / 55 | ↓47.1 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #29 / 51 | ↓88.5 | Source ↗official | |
| CAIS Risk Indexmask | #33 / 57 | ↓23.7 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #49 / 51 | ↓63.3 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #11 / 54 | ↓15.2 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #24 / 104 | ↓9.521 | Source ↗official | |
| DelusionEvaldelusional_prevalence_pct | #4 / 16 | ↓18 | Source ↗official | |
| DelusionEvaldiscourages_harm_prevalence_pct | #3 / 16 | ↑52.2 | Source ↗official | |
| DelusionEvalfacilitates_harm_prevalence_pct | #1 / 16 | ↓0 | Source ↗official | |
| DelusionEvalrelationship_prevalence_pct | #4 / 16 | ↓11.8 | Source ↗official | |
| DelusionEvalsycophancy_prevalence_pct | #3 / 16 | ↓15.6 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #24 / 50 | ↓88.5 | Source ↗official | |
| SM-Benchadversarial | #92 / 92 | ↑68.29 | Source ↗official | |
| SM-Benchambiguous_interpretation | #44 / 92 | ↑86.31 | Source ↗official | |
| SM-Benchanti_hallucination | #89 / 92 | ↑68.06 | Source ↗official | |
| SM-Bencheq_boundaries | #65 / 92 | ↑58.71 | Source ↗official | |
| SM-Benchoverfit | #92 / 92 | ↑4.37 | Source ↗official | |
| Vals AI Cheating Auditterminal_bench_cheating_shortcut_evidence_rate_pct | #4 / 14 | ↓0.7491 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #2 / 94 | ↑96.9 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #25 / 25 | ↓18.9 | 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 |
|---|---|---|---|---|
| CAIS Risk Indexbioweapons_assistance | #17 / 54 | ↑78.3 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #80 / 104 | ↑9.521 | Source ↗official | |
| DelusionEvaldiscourages_harm_prevalence_pct | #14 / 16 | ↓52.2 | Source ↗official | |
| DelusionEvalfacilitates_harm_prevalence_pct | #15 / 16 | ↑0 | Source ↗official | |
| SM-Benchadversarial | #1 / 92 | ↓68.29 | Source ↗official | |
| SM-Bencheq_boundaries | #65 / 92 | ↑58.71 | Source ↗official | |
| SM-Benchoverfit | #92 / 92 | ↑4.37 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #167 / 181 | ↑32.2 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #25 / 25 | ↓18.9 | Source ↗official |