Model profile
Grok 4.3
Evidence summary
Grok 4.3 has an estimated overall rank of #104; its 90% source-sensitivity interval is #61–#204. Its behavior-only rank is #94; company governance moves the combined estimate to #104. Published evidence spans 20 evals and 7 of 7 behavior components. Its strongest relative result is AA-Omniscience (hallucination_rate, #5 of 328); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #236 of 241).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-4-3
- OpenRouter ↗x-ai/grok-4.3
- Official model documentation ↗Family-level model document · xai · first party
- Release source ↗curated outward evidence
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 | #5 / 328 | ↓0.1694 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #9 / 30 | ↑1211.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #20 / 112 | ↑1458.0 | Source ↗official | |
| BioSecBench-Refusalbalanced_refusal_score | #9 / 10 | ↑0.1092 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #25 / 106 | ↑0.48 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #33 / 44 | ↓90.1 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #14 / 49 | ↓60.1 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #10 / 50 | ↓42.5 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #16 / 46 | ↓84.7 | Source ↗official | |
| CAIS Risk Indexmask | #40 / 52 | ↓48.3 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #23 / 33 | ↓51.6 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #2 / 49 | ↓3.9 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #46 / 104 | ↓32.95 | Source ↗official | |
| DystopiaBenchbasaglia_score | #30 / 50 | ↓65.37 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #29 / 50 | ↓61.77 | Source ↗official | |
| DystopiaBenchhuxley_score | #25 / 50 | ↓71.63 | Source ↗official | |
| DystopiaBenchlaguardia_score | #45 / 50 | ↓71.3 | Source ↗official | |
| DystopiaBenchorwell_score | #30 / 50 | ↓72.07 | Source ↗official | |
| DystopiaBenchpetrov_score | #23 / 50 | ↓72.07 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #165 / 241 | ↑12.92 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #236 / 241 | ↑65.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #30 / 241 | ↑95 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #190 / 239 | ↑91.45 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #16 / 50 | ↓84.7 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #13 / 31 | ↓8.2 | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #12 / 13 | ↓67 | Source ↗official | |
| MANTAAWMS | #6 / 7 | ↑0.371 | Source ↗official | |
| MANTAAWVS | #6 / 7 | ↑0.352 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #29 / 66 | ↑0.47 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #23 / 70 | ↑0.7946 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #36 / 70 | ↑0.9355 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #28 / 67 | ↑0.545 | Source ↗official | |
| SM-Benchadversarial | #30 / 79 | ↑83.41 | Source ↗official | |
| SM-Benchambiguous_interpretation | #37 / 79 | ↑86.01 | Source ↗official | |
| SM-Benchanti_hallucination | #18 / 79 | ↑97.91 | Source ↗official | |
| SM-Bencheq_boundaries | #15 / 79 | ↑70.79 | Source ↗official | |
| SM-Benchoverfit | #26 / 79 | ↑80.87 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #29 / 113 | ↑6.93 | Source ↗official | |
| SpeciEvalland_animal_4ns | #61 / 113 | ↓4.58 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #36 / 113 | ↓4.65 | Source ↗official | |
| SpeciEvalspeciesism | #77 / 113 | ↓2.25 | Source ↗official | |
| TukaBenchafri_jbb_cultural_asr | #1 / 6 | ↓10.6 | Source ↗official | |
| TukaBenchafri_jbb_harm_asr | #2 / 6 | ↓8.2 | Source ↗official | |
| TukaBenchafrijail_mono_asr | #2 / 6 | ↓17.7 | 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.