Model profile
O1 Preview
OpenAIdeveloper
2024-09-12release date
#41 / 309overall rank
4eval lineages
Evidence summary
O1 Preview has an estimated overall rank of #41; its 90% source-sensitivity interval is #6–#157. Its behavior-only rank is #44; company governance moves the combined estimate to #41. Published evidence spans 4 evals and 6 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #14 of 241); its weakest is SORRY-Bench (avg, #22 of 51).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗o1-preview
- OpenRouter ↗openai/o1-preview
- System card ↗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 | Source |
|---|---|---|---|---|
| Confabulationsconfabulation_rate | #21 / 52 | ↓18.32 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #64 / 241 | ↑23.51 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #14 / 241 | ↑95.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #17 / 241 | ↑98.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #86 / 239 | ↑97.64 | Source ↗official | |
| Large-scale Moral Machine experiment on LLMshuman_choice_distance | #11 / 39 | ↓0.8621 | Source ↗official | |
| SORRY-Benchavg | #22 / 51 | ↓0.25 | Source ↗official |