Model profile
GPT 5.6 Sol
OpenAIdeveloper
2026-07-09release date
#32 / 267overall rank
17eval lineages
1discovery sources
Evidence summary
GPT 5.6 Sol has an estimated overall rank of #32; its 90% source-sensitivity interval is #6–#175. Its behavior-only rank is #39; company governance moves the combined estimate to #32. Published evidence spans 17 evals and 7 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is CAIS Risk Index (textquests_harm, #48 of 48).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-5-6-sol
- OpenRouter ↗openai/gpt-5.6-sol
- Official model documentation ↗Family-level 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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #220 / 311 | 0.8711 | ↓ lower | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #26 / 105 | 0.465 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #2 / 43 | 40.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #21 / 48 | 64.3 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #15 / 49 | 46.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #9 / 45 | 82.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #3 / 51 | 4.6 | ↓ lower | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #10 / 32 | 41.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #48 / 48 | 31.4 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #53 / 260 | 27.91 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #214 / 260 | 80.83 | ↑ 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 | #211 / 258 | 91.09 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsextremism_not_unsafe | #6 / 7 | 0.962 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsgore_not_unsafe | #5 / 7 | 0.708 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptshate_not_unsafe | #4 / 7 | 0.982 | ↑ 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 | #5 / 7 | 0.945 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptssexual_minors_not_unsafe | #3 / 7 | 0.973 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptssexual_not_unsafe | #2 / 7 | 0.948 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — disallowed content with challenging promptsviolent_illicit_not_unsafe | #7 / 7 | 0.934 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — first-person fairnessharm_overall_pct | #4 / 7 | 0.98 | ↓ lower | Source ↗official | |
| GPT-5.6 system card — prompt-injection robustnessconnectors_injection_resistance | #1 / 7 | 1 | ↑ higher | Source ↗official | |
| GPT-5.6 system card — prompt-injection robustnesssearch_function_calling_injection_resistance | #2 / 6 | 0.91 | ↑ higher | Source ↗official | |
| Gray Swan indirect prompt injection (15 attempts)attack_success_probability_k15_pct | #7 / 13 | 20 | ↓ lower | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #35 / 54 | 26.14 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #9 / 50 | 82.7 | ↓ lower | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #22 / 31 | 8.9 | ↓ lower | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #1 / 13 | 0 | ↓ lower | Source ↗official | |
| SM-Benchadversarial | #26 / 73 | 83.66 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #39 / 73 | 84.08 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #36 / 73 | 92.41 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #33 / 73 | 65.31 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #51 / 73 | 57.52 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #25 / 102 | 4.315 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #15 / 102 | 4.435 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #3 / 102 | 1.165 | ↓ lower | Source ↗official | |
| TACbase_welfare_rate | #65 / 68 | 16.67 | ↑ higher | 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.