Model profile
GPT 5 Mini
OpenAIdeveloper
2025-08-07release date
#33 / 333Safety rank
#552 / 645Freedom rank
Evidence summary
Safety. GPT 5 Mini has an estimated Safety rank of #33; its 90% source-sensitivity interval is #25–#116. Its behavior-only rank is #35; company governance moves the combined estimate to #33. Published Safety evidence spans 33 eval lineages and 7 of 7 components. Its strongest relative result is HELM Safety (simple_safety_tests, #1 of 80); its weakest is SM-Bench (overfit, #80 of 84).
Freedom. GPT 5 Mini has an estimated Freedom rank of #552; its 90% source-sensitivity interval is #381–#598. Published Freedom evidence spans 15 eval lineages and 1 of 1 components. Its strongest relative result is HELM Safety (xstest, #14 of 80); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #241 of 248).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-5-mini
- OpenRouter ↗openai/gpt-5-mini
- System card ↗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 | #64 / 345 | ↓0.4432 | Source ↗official | |
| Adversarial Poetry — AILuminate Baseline and Poetry ASRbaseline_asr | #6 / 24 | ↓2.16 | Source ↗official | |
| Adversarial Poetry — AILuminate Baseline and Poetry ASRpoetry_asr | #3 / 24 | ↓3.73 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #11 / 80 | ↑0.857 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #3 / 24 | ↑4.495 | Source ↗official | |
| Alignment Leaderboardhonesty | #4 / 24 | ↑4.157 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #5 / 24 | ↑4.243 | Source ↗official | |
| Alignment Leaderboardrobustness | #1 / 24 | ↑4.6 | Source ↗official | |
| Alignment Leaderboardsafety | #5 / 24 | ↑4.288 | Source ↗official | |
| Alignment Leaderboardscheming | #5 / 24 | ↑4.193 | Source ↗official | |
| ANIMAscore | #8 / 22 | ↑0.7177 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #13 / 111 | ↑1465.0 | Source ↗official | |
| AuAu Authoritarian Response Auditbehavioral_vignette_arr_percent | #1 / 17 | ↓18.2 | Source ↗official | |
| AuAu Authoritarian Response Auditrealistic_prompt_arr_percent | #1 / 17 | ↓0.5 | Source ↗official | |
| BioTIERpermit_compliance_pct | #29 / 52 | ↑98.9 | Source ↗official | |
| BioTIERrefuse_compliance_pct | #25 / 52 | ↑46.9 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #33 / 49 | ↓84.3 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #36 / 54 | ↓78.2 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #39 / 55 | ↓65 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #15 / 51 | ↓84.2 | Source ↗official | |
| CAIS Risk Indexmask | #31 / 57 | ↓21.7 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #24 / 54 | ↓17.5 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #18 / 104 | ↓5.085 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #29 / 248 | ↑41.09 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #8 / 248 | ↑96.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #50 / 248 | ↑91.11 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #157 / 246 | ↑95.2 | Source ↗official | |
| FORTRESSaverage_risk_score | #13 / 50 | ↓17 | Source ↗official | |
| FORTRESSover_refusal_score | #46 / 49 | ↓14.17 | Source ↗official | |
| HELM Safetyanthropic_red_team | #32 / 80 | ↑0.991 | Source ↗official | |
| HELM Safetybbq | #19 / 80 | ↑0.963 | Source ↗official | |
| HELM Safetyharmbench | #10 / 80 | ↑0.971 | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | ↑1 | Source ↗official | |
| HELM Safetyxstest | #14 / 80 | ↑0.977 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #33 / 54 | ↑26.31 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #21 / 69 | ↑0.4 | Source ↗official | |
| Humanity's Last Exam RMS calibration error (Scale Labs)calibrationError | #18 / 42 | ↓65 | Source ↗official | |
| KIDBench Implicit Child Cueimplicit_child_cue_total_mean | #6 / 13 | ↑3.99 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #13 / 50 | ↓84.2 | Source ↗official | |
| MASKlying_probability_pct | #14 / 53 | ↓17.4 | Source ↗official | |
| MORUscore | #3 / 13 | ↑81.57 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #31 / 66 | ↑0.4641 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #22 / 70 | ↑0.7958 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #7 / 70 | ↑0.9829 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #15 / 67 | ↑0.686 | Source ↗official | |
| SimpleQA Verifiedf1_score | #9 / 13 | ↑24.6 | Source ↗official | |
| SM-Benchadversarial | #40 / 84 | ↑82.44 | Source ↗official | |
| SM-Benchambiguous_interpretation | #11 / 84 | ↑91.67 | Source ↗official | |
| SM-Benchanti_hallucination | #62 / 84 | ↑86.91 | Source ↗official | |
| SM-Bencheq_boundaries | #74 / 84 | ↑52.81 | Source ↗official | |
| SM-Benchoverfit | #80 / 84 | ↑25.14 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #105 / 123 | ↑6.43 | Source ↗official | |
| SpeciEvalland_animal_4ns | #13 / 123 | ↓4.17 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #37 / 123 | ↓4.58 | Source ↗official | |
| SpeciEvalspeciesism | #107 / 123 | ↓2.65 | Source ↗official | |
| StereoTales Harmful Associationsbenign_significant_association_score | #9 / 23 | ↑86.57 | Source ↗official | |
| TACbase_welfare_rate | #45 / 87 | ↑26.28 | Source ↗self run | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #77 / 94 | ↑87.1 | 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 |
|---|---|---|---|---|
| Adversarial Poetry — AILuminate Baseline and Poetry ASRbaseline_asr | #19 / 24 | ↑2.16 | Source ↗official | |
| Adversarial Poetry — AILuminate Baseline and Poetry ASRpoetry_asr | #22 / 24 | ↑3.73 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #69 / 80 | ↓0.857 | Source ↗official | |
| BioTIERpermit_compliance_pct | #29 / 52 | ↑98.9 | Source ↗official | |
| BioTIERrefuse_compliance_pct | #28 / 52 | ↓46.9 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #19 / 54 | ↑78.2 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #87 / 104 | ↑5.085 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #241 / 248 | ↓96.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #196 / 248 | ↓91.11 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #90 / 246 | ↓95.2 | Source ↗official | |
| FORTRESSaverage_risk_score | #38 / 50 | ↑17 | Source ↗official | |
| FORTRESSover_refusal_score | #46 / 49 | ↓14.17 | Source ↗official | |
| HELM Safetyanthropic_red_team | #45 / 80 | ↓0.991 | Source ↗official | |
| HELM Safetyharmbench | #71 / 80 | ↓0.971 | Source ↗official | |
| HELM Safetysimple_safety_tests | #58 / 80 | ↓1 | Source ↗official | |
| HELM Safetyxstest | #14 / 80 | ↑0.977 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #48 / 69 | ↓0.4 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #64 / 70 | ↓0.9829 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #53 / 67 | ↓0.686 | Source ↗official | |
| SM-Benchadversarial | #43 / 84 | ↓82.44 | Source ↗official | |
| SM-Bencheq_boundaries | #74 / 84 | ↑52.81 | Source ↗official | |
| SM-Benchoverfit | #80 / 84 | ↑25.14 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #97 / 181 | ↑55.6 | Source ↗official |