Model profile
GPT 5.3 Codex
Evidence summary
Safety. GPT 5.3 Codex has an estimated Safety rank of #103; its 90% source-sensitivity interval is #58–#255. Its behavior-only rank is #114; company governance moves the combined estimate to #103. Published Safety evidence spans 5 eval lineages and 5 of 7 components. Its strongest relative result is DystopiaBench (petrov_score, #13 of 50); its weakest is OpenAI GPT-5.4 Property Preservation (user_work_preserved, #3 of 3).
Freedom. GPT 5.3 Codex has an estimated Freedom rank of #381; its 90% source-sensitivity interval is #186–#475. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (adversarial, #20 of 84); its weakest is SM-Bench (overfit, #75 of 84).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗gpt-5-3-codex
- OpenRouter ↗openai/gpt-5.3-codex
- Official model documentation ↗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 | #261 / 345 | ↓0.8918 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #73 / 117 | ↑0.21 | Source ↗official | |
| DystopiaBenchbasaglia_score | #21 / 50 | ↓59.4 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #18 / 50 | ↓45.47 | Source ↗official | |
| DystopiaBenchhuxley_score | #16 / 50 | ↓53.17 | Source ↗official | |
| DystopiaBenchlaguardia_score | #20 / 50 | ↓62.37 | Source ↗official | |
| DystopiaBenchorwell_score | #20 / 50 | ↓57.3 | Source ↗official | |
| DystopiaBenchpetrov_score | #13 / 50 | ↓50.6 | Source ↗official | |
| OpenAI GPT-5.4 Property Preservationdestructive_action_avoidance | #1 / 3 | ↑0.88 | Source ↗official | |
| OpenAI GPT-5.4 Property Preservationuser_work_preserved | #3 / 3 | ↑0.08 | Source ↗official | |
| OpenAI GPT-5.4 User Confirmationsfinancial_transaction | #3 / 3 | ↑0.99 | Source ↗official | |
| OpenAI GPT-5.4 User Confirmationsgeneral_confirmation | #3 / 3 | ↑0.91 | Source ↗official | |
| OpenAI GPT-5.4 User Confirmationshigh_stakes_communication | #3 / 3 | ↑0.99 | Source ↗official | |
| SM-Benchadversarial | #63 / 84 | ↑79.02 | Source ↗official | |
| SM-Benchambiguous_interpretation | #37 / 84 | ↑86.9 | Source ↗official | |
| SM-Benchanti_hallucination | #43 / 84 | ↑93.72 | Source ↗official | |
| SM-Bencheq_boundaries | #22 / 84 | ↑69.1 | Source ↗official | |
| SM-Benchoverfit | #75 / 84 | ↑33.33 | 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.