Model profile
GPT-5.3 Chat
OpenAIdeveloper
2026-03-03release date
#9 / 309overall rank
6eval lineages
Evidence summary
GPT-5.3 Chat has an estimated overall rank of #9; its 90% source-sensitivity interval is #2–#74. Its behavior-only rank is #12; company governance moves the combined estimate to #9. Published evidence spans 6 evals and 5 of 7 behavior components. Its strongest relative result is SpeciEval (land_animal_4ns, #1 of 113); its weakest is SpeciEval (speciesism, #106 of 113).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- OpenRouter ↗openai/gpt-5.3-chat
- Official model documentation ↗Family-level model document · openai · first party
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 |
|---|---|---|---|---|
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #6 / 112 | ↑1485.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #37 / 106 | ↑0.4 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #15 / 104 | ↓4.337 | Source ↗official | |
| SM-Benchadversarial | #49 / 79 | ↑80 | Source ↗official | |
| SM-Benchambiguous_interpretation | #26 / 79 | ↑88.69 | Source ↗official | |
| SM-Benchanti_hallucination | #47 / 79 | ↑90.58 | Source ↗official | |
| SM-Bencheq_boundaries | #40 / 79 | ↑64.89 | Source ↗official | |
| SM-Benchoverfit | #67 / 79 | ↑41.53 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #50 / 113 | ↑6.83 | Source ↗official | |
| SpeciEvalland_animal_4ns | #1 / 113 | ↓3.52 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #2 / 113 | ↓3.73 | Source ↗official | |
| SpeciEvalspeciesism | #106 / 113 | ↓3 | Source ↗official | |
| Vigil Mental Health Safetyoverall_score | #6 / 23 | ↑71 | Source ↗official |