Model profile
GPT 5.5
OpenAIdeveloper
2026-04-23release date
#9 / 267overall rank
29eval lineages
2discovery sources
Evidence summary
GPT 5.5 has an estimated overall rank of #9; its 90% source-sensitivity interval is #10–#66. Its behavior-only rank is #10; company governance moves the combined estimate to #9. Published evidence spans 29 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is GPT-5.6 system card — disallowed content with challenging prompts (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
- openai.com ↗
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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #205 / 311 | 0.8553 | ↓ lower | Source ↗official | |
| AgentAbstainabstain | #9 / 17 | 61.1 | ↑ higher | Source ↗official | |
| AgentAbstaincar | #7 / 17 | 59.8 | ↑ higher | Source ↗official | |
| AgentAbstainpaired | #4 / 17 | 52.5 | ↑ higher | Source ↗official | |
| ANIMAscore | #3 / 19 | 0.74 | ↑ higher | Source ↗official | |
| BioSecBench-Refusalbalanced_refusal_score | #8 / 10 | 0.2869 | ↑ higher | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #28 / 105 | 0.4567 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #3 / 43 | 41.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #14 / 48 | 60.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #7 / 49 | 41.8 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #8 / 45 | 81.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #12 / 51 | 9.9 | ↓ lower | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #6 / 32 | 39 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #36 / 48 | 21.4 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #5 / 105 | 2.343 | ↓ lower | Source ↗official | |
| DystopiaBenchbasaglia_score | #17 / 50 | 55 | ↓ lower | Source ↗official | |
| DystopiaBenchbaudrillard_score | #14 / 50 | 40.7 | ↓ lower | Source ↗official | |
| DystopiaBenchhuxley_score | #15 / 50 | 52.5 | ↓ lower | Source ↗official | |
| DystopiaBenchlaguardia_score | #17 / 50 | 60.13 | ↓ lower | Source ↗official | |
| DystopiaBenchorwell_score | #15 / 50 | 49.5 | ↓ lower | Source ↗official | |
| DystopiaBenchpetrov_score | #12 / 50 | 49.13 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #40 / 260 | 38.5 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #219 / 260 | 79.33 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #7 / 260 | 99.44 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #201 / 258 | 92 | ↑ higher | Source ↗official | |
| FORTRESSaverage_risk_score | #10 / 49 | 16.28 | ↓ lower | Source ↗official | |
| FORTRESSover_refusal_score | #38 / 46 | 9.79 | ↓ lower | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsextremism_not_unsafe | #7 / 7 | 0.925 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsgore_not_unsafe | #3 / 7 | 0.8 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptshate_not_unsafe | #1 / 7 | 1 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsnonviolent_illicit_not_unsafe | #6 / 7 | 0.987 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsself_harm_not_unsafe | #7 / 7 | 0.917 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptssexual_minors_not_unsafe | #6 / 7 | 0.938 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptssexual_not_unsafe | #3 / 7 | 0.944 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsviolent_illicit_not_unsafe | #5 / 7 | 0.94 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — first-person fairnessharm_overall_pct | #6 / 7 | 1.12 | ↓ lower | Source ↗official | |
| GPT-5.6 system card — prompt-injection robustnessconnectors_injection_resistance | #1 / 7 | 1 | ↑ higher | Source ↗official | |
| Gray Swan indirect prompt injection (15 attempts)attack_success_probability_k15_pct | #8 / 13 | 20.8 | ↓ lower | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #11 / 54 | 28.38 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #8 / 50 | 81.7 | ↓ lower | Source ↗official | |
| MANTAAWMS | #2 / 7 | 0.504 | ↑ higher | Source ↗official | |
| MANTAAWVS | #2 / 7 | 0.664 | ↑ higher | Source ↗official | |
| MORUscore | #1 / 13 | 85.4 | ↑ higher | Source ↗official | |
| ODCV-Benchaverage_severity | #2 / 12 | 0.7125 | ↓ lower | Source ↗official | |
| ODCV-Benchmisalignment_rate | #3 / 12 | 21.25 | ↓ lower | Source ↗official | |
| PHAREbias_resistance_diagnostic | #48 / 66 | 0.3829 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #20 / 70 | 0.7978 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #6 / 70 | 0.9892 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #18 / 67 | 0.6513 | ↑ higher | Source ↗official | |
| RefusalBenchyouden_j | #11 / 19 | 0.4213 | ↑ higher | Source ↗official | |
| SM-Benchadversarial | #31 / 73 | 82.93 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #42 / 73 | 82.74 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #27 / 73 | 94.76 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #36 / 73 | 64.61 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #57 / 73 | 50 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #54 / 102 | 4.58 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #24 / 102 | 4.55 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #7 / 102 | 1.32 | ↓ lower | Source ↗official | |
| TACbase_welfare_rate | #57 / 68 | 19.23 | ↑ higher | Source ↗official | |
| ToolPrivacyBenchprivate_mt_poi | #3 / 9 | 20.39 | ↓ lower | Source ↗official | |
| ToolPrivacyBenchpublic_mt_poi | #2 / 9 | 16.75 | ↓ lower | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #9 / 25 | 7.5 | ↓ lower | 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.