Model profile
Grok 4.5
Evidence summary
Grok 4.5 has an estimated overall rank of #65; its 90% source-sensitivity interval is #27–#190. Its behavior-only rank is #56; company governance moves the combined estimate to #65. Published evidence spans 13 evals and 6 of 7 behavior components. Its strongest relative result is SpeciEval (belief_animal_sentience, #1 of 102); its weakest is Gray Swan indirect prompt injection (15 attempts) (attack_success_probability_k15_pct, #13 of 13).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-4-5
- OpenRouter ↗x-ai/grok-4.5
- Official model documentation ↗Family-level model document · xai · 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 | Better | Source |
|---|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #59 / 311 | 0.5353 | ↓ lower | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #20 / 105 | 0.54 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #26 / 43 | 83.5 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #33 / 48 | 79.2 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #5 / 49 | 30.8 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #29 / 45 | 90.9 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #31 / 51 | 27.9 | ↓ lower | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #20 / 32 | 50.2 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #5 / 48 | 11.2 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #40 / 105 | 22.98 | ↓ lower | Source ↗official | |
| Gray Swan indirect prompt injection (15 attempts)attack_success_probability_k15_pct | #13 / 13 | 60.8 | ↓ lower | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #30 / 50 | 90.9 | ↓ lower | Source ↗official | |
| SM-Benchadversarial | #20 / 73 | 85.37 | ↑ higher | Source ↗official | |
| SM-Benchambiguous_interpretation | #6 / 73 | 91.96 | ↑ higher | Source ↗official | |
| SM-Benchanti_hallucination | #15 / 73 | 97.91 | ↑ higher | Source ↗official | |
| SM-Bencheq_boundaries | #18 / 73 | 69.1 | ↑ higher | Source ↗official | |
| SM-Benchoverfit | #7 / 73 | 92.35 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 102 | 7 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #98 / 102 | 5.15 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #92 / 102 | 5.15 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #100 / 102 | 3.52 | ↓ lower | 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.