Model profile
O1
Evidence summary
O1 has an estimated overall rank of #108; its 90% source-sensitivity interval is #52–#180. Its behavior-only rank is #114; company governance moves the combined estimate to #108. Published evidence spans 17 evals and 6 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #9 of 241); its weakest is CAIS Risk Index (hle_overconfidence, #48 of 50).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗o1
- OpenRouter ↗openai/o1
- 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 |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #109 / 328 | ↓0.6963 | Source ↗official | |
| AbstentionBenchanswer_unknown_f1 | #7 / 20 | ↑0.8917 | Source ↗official | |
| AbstentionBenchfalse_premise_f1 | #1 / 20 | ↑0.7687 | Source ↗official | |
| AbstentionBenchstale_f1 | #8 / 20 | ↑0.646 | Source ↗official | |
| AbstentionBenchsubjective_f1 | #5 / 20 | ↑0.7654 | Source ↗official | |
| AbstentionBenchunderspecified_context_f1 | #6 / 20 | ↑0.6907 | Source ↗official | |
| AbstentionBenchunderspecified_intent_f1 | #3 / 20 | ↑0.7694 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #21 / 80 | ↑0.8 | Source ↗official | |
| BlueBench AttaQ-100attaq_harmlessness_reward_pct | #15 / 18 | ↑82.96 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #26 / 49 | ↓68.1 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #48 / 50 | ↓83 | Source ↗official | |
| CAIS Risk Indexmask | #34 / 52 | ↓40.7 | Source ↗official | |
| Confabulationsconfabulation_rate | #10 / 52 | ↓10.89 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #102 / 241 | ↑17.83 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #9 / 241 | ↑96.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #27 / 241 | ↑96.11 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #92 / 239 | ↑97.41 | Source ↗official | |
| FORTRESSaverage_risk_score | #17 / 49 | ↓19.38 | Source ↗official | |
| FORTRESSover_refusal_score | #25 / 46 | ↓5.05 | Source ↗official | |
| HELM Safetyanthropic_red_team | #50 / 80 | ↑0.983 | Source ↗official | |
| HELM Safetybbq | #9 / 80 | ↑0.973 | Source ↗official | |
| HELM Safetyharmbench | #12 / 80 | ↑0.963 | Source ↗official | |
| HELM Safetysimple_safety_tests | #32 / 80 | ↑0.99 | Source ↗official | |
| HELM Safetyxstest | #24 / 80 | ↑0.97 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #45 / 54 | ↑24.33 | Source ↗official | |
| MASKlying_probability_pct | #28 / 53 | ↓40.73 | Source ↗official | |
| Reward Hacking Benchmarkintegrity_score | #9 / 13 | ↑93.2 | 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.