Model profile
GPT 5.5
OpenAIdeveloper
2026-04-23release date
#14 / 312overall rank
38eval lineages
Evidence summary
GPT 5.5 has an estimated overall rank of #14; its 90% source-sensitivity interval is #11–#79. Its behavior-only rank is #15; company governance moves the combined estimate to #14. Published evidence spans 38 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 113); its weakest is GPT-5.6 system card (extremism_not_unsafe, #7 of 7).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-5-5
- OpenRouter ↗openai/gpt-5.5
- 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 | #231 / 330 | ↓0.8804 | Source ↗official | |
| AgentAbstainabstain | #9 / 17 | ↑61.1 | Source ↗official | |
| AgentAbstaincar | #7 / 17 | ↑59.8 | Source ↗official | |
| AgentAbstainpaired | #4 / 17 | ↑52.5 | Source ↗official | |
| ANIMAscore | #4 / 22 | ↑0.74 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #2 / 30 | ↑1269.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #4 / 112 | ↑1490.0 | Source ↗official | |
| BioSecBench-Refusalbalanced_refusal_score | #8 / 10 | ↑0.2869 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #28 / 106 | ↑0.4567 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #4 / 45 | ↓41.5 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #15 / 50 | ↓60.5 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #8 / 51 | ↓41.8 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #8 / 47 | ↓81.7 | Source ↗official | |
| CAIS Risk Indexmask | #13 / 53 | ↓9.9 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #10 / 34 | ↓40 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #36 / 50 | ↓21.4 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #5 / 104 | ↓2.343 | Source ↗official | |
| DystopiaBenchbasaglia_score | #17 / 50 | ↓55 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #14 / 50 | ↓40.7 | Source ↗official | |
| DystopiaBenchhuxley_score | #15 / 50 | ↓52.5 | Source ↗official | |
| DystopiaBenchlaguardia_score | #17 / 50 | ↓60.13 | Source ↗official | |
| DystopiaBenchorwell_score | #15 / 50 | ↓49.5 | Source ↗official | |
| DystopiaBenchpetrov_score | #12 / 50 | ↓49.13 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #37 / 241 | ↑38.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #200 / 241 | ↑79.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #6 / 241 | ↑99.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #184 / 239 | ↑92 | Source ↗official | |
| FORTRESSaverage_risk_score | #10 / 49 | ↓16.28 | Source ↗official | |
| FORTRESSover_refusal_score | #40 / 48 | ↓9.79 | Source ↗official | |
| GPT-5.6 system cardconnectors_injection_resistance | #1 / 7 | ↑1 | Source ↗official | |
| GPT-5.6 system cardemotional_reliance | #6 / 7 | ↑0.915 | Source ↗official | |
| GPT-5.6 system cardextremism_not_unsafe | #7 / 7 | ↑0.925 | Source ↗official | |
| GPT-5.6 system cardgore_not_unsafe | #3 / 7 | ↑0.8 | Source ↗official | |
| GPT-5.6 system cardharm_overall_pct | #6 / 7 | ↓1.12 | Source ↗official | |
| GPT-5.6 system cardhate_not_unsafe | #1 / 7 | ↑1 | Source ↗official | |
| GPT-5.6 system cardmental_health | #6 / 7 | ↑0.82 | Source ↗official | |
| GPT-5.6 system cardnonviolent_illicit_not_unsafe | #6 / 7 | ↑0.987 | Source ↗official | |
| GPT-5.6 system cardself_harm | #6 / 7 | ↑0.868 | Source ↗official | |
| GPT-5.6 system cardself_harm_not_unsafe | #7 / 7 | ↑0.917 | Source ↗official | |
| GPT-5.6 system cardsexual_minors_not_unsafe | #6 / 7 | ↑0.938 | Source ↗official | |
| GPT-5.6 system cardsexual_not_unsafe | #3 / 7 | ↑0.944 | Source ↗official | |
| GPT-5.6 system cardviolent_illicit_not_unsafe | #5 / 7 | ↑0.94 | Source ↗official | |
| Gray Swan indirect prompt injection (15 attempts)attack_success_probability_k15_pct | #8 / 13 | ↓20.8 | Source ↗official | |
| HarmVideoBenchharmful_video_safety_recognition_reasoning | #3 / 19 | ↑0.809 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #11 / 54 | ↑28.38 | Source ↗official | |
| JuICE Cultural-Error Span Detectionf1 | #2 / 10 | ↑0.5112 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingemotional_safety | #4 / 10 | ↑82.4 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingidentity_collapse | #2 / 10 | ↑92.6 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingsycophancy_spine | #7 / 10 | ↑82.4 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingvalue_integrity | #6 / 10 | ↑90.5 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #8 / 50 | ↓81.7 | Source ↗official | |
| MANTAAWMS | #2 / 7 | ↑0.504 | Source ↗official | |
| MANTAAWVS | #2 / 7 | ↑0.664 | Source ↗official | |
| MORUscore | #1 / 13 | ↑85.4 | Source ↗official | |
| MuPPET Contextual Privacymultiparty_contextual_privacy_score | #2 / 7 | ↑50.98 | Source ↗official | |
| ODCV-Benchaverage_severity | #2 / 12 | ↓0.7125 | Source ↗official | |
| ODCV-Benchmisalignment_rate | #3 / 12 | ↓21.25 | Source ↗official | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #3 / 19 | ↓2 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #48 / 66 | ↑0.3829 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #20 / 70 | ↑0.7978 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #6 / 70 | ↑0.9892 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #18 / 67 | ↑0.6513 | Source ↗official | |
| RefusalBenchyouden_j | #6 / 19 | ↑0.2979 | Source ↗official | |
| SM-Benchadversarial | #33 / 79 | ↑82.93 | Source ↗official | |
| SM-Benchambiguous_interpretation | #47 / 79 | ↑82.74 | Source ↗official | |
| SM-Benchanti_hallucination | #31 / 79 | ↑94.76 | Source ↗official | |
| SM-Bencheq_boundaries | #41 / 79 | ↑64.61 | Source ↗official | |
| SM-Benchoverfit | #62 / 79 | ↑50 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 113 | ↑7 | Source ↗official | |
| SpeciEvalland_animal_4ns | #61 / 113 | ↓4.58 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #27 / 113 | ↓4.55 | Source ↗official | |
| SpeciEvalspeciesism | #10 / 113 | ↓1.32 | Source ↗official | |
| TACbase_welfare_rate | #62 / 76 | ↑19.23 | Source ↗official | |
| ToolPrivacyBenchprivate_mt_poi | #3 / 9 | ↓20.39 | Source ↗official | |
| ToolPrivacyBenchpublic_mt_poi | #2 / 9 | ↓16.75 | Source ↗official | |
| UK AISI cyber-evaluation cheating and prompted self-reportattempted_cheating_trajectory_rate_pct | #3 / 5 | ↓11.4 | Source ↗official | |
| UK AISI cyber-evaluation cheating and prompted self-reportspecific_cheating_action_mention_rate_pct | #4 / 5 | ↑75 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #41 / 94 | ↑90.7 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #9 / 25 | ↓7.5 | 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.