Model profile
Grok 4.20
xAIdeveloper
2026-03-10release date
#121 / 312overall rank
22eval lineages
Evidence summary
Grok 4.20 has an estimated overall rank of #121; its 90% source-sensitivity interval is #60–#240. Its behavior-only rank is #116; company governance moves the combined estimate to #121. Published evidence spans 22 evals and 7 of 7 behavior components. Its strongest relative result is SM-Bench (eq_boundaries, #1 of 79); its weakest is BioSecBench-Refusal (balanced_refusal_score, #10 of 10).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-4-20
- OpenRouter ↗x-ai/grok-4.20
- 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 | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #7 / 330 | ↓0.1992 | Source ↗official | |
| Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating | #10 / 30 | ↑1210.0 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #37 / 112 | ↑1446.0 | Source ↗official | |
| BioSecBench-Refusalbalanced_refusal_score | #10 / 10 | ↑0.02854 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #18 / 106 | ↑0.55 | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #30 / 45 | ↓84.6 | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #11 / 50 | ↓56.9 | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #14 / 51 | ↓45.8 | Source ↗official | |
| CAIS Risk Indexmachiavelli | #11 / 47 | ↓83.8 | Source ↗official | |
| CAIS Risk Indexmask | #35 / 53 | ↓41 | Source ↗official | |
| CAIS Risk Indexpolitical_manipulation | #9 / 34 | ↓39.9 | Source ↗official | |
| CAIS Risk Indextextquests_harm | #4 / 50 | ↓9.9 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #45 / 104 | ↓31.95 | Source ↗official | |
| DelusionEvaldelusional_prevalence_pct | #16 / 16 | ↓70 | Source ↗official | |
| DelusionEvaldiscourages_harm_prevalence_pct | #6 / 16 | ↑31.4 | Source ↗official | |
| DelusionEvalfacilitates_harm_prevalence_pct | #16 / 16 | ↓11.6 | Source ↗official | |
| DelusionEvalrelationship_prevalence_pct | #16 / 16 | ↓53.4 | Source ↗official | |
| DelusionEvalsycophancy_prevalence_pct | #15 / 16 | ↓36.2 | Source ↗official | |
| DystopiaBenchbasaglia_score | #44 / 50 | ↓70.5 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #41 / 50 | ↓67.37 | Source ↗official | |
| DystopiaBenchhuxley_score | #36 / 50 | ↓75.07 | Source ↗official | |
| DystopiaBenchlaguardia_score | #42 / 50 | ↓69.8 | Source ↗official | |
| DystopiaBenchorwell_score | #42 / 50 | ↓75.07 | Source ↗official | |
| DystopiaBenchpetrov_score | #33 / 50 | ↓75.03 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #26 / 54 | ↑27.15 | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #11 / 50 | ↓83.8 | Source ↗official | |
| ODCV-Benchaverage_severity | #12 / 12 | ↓2.564 | Source ↗official | |
| ODCV-Benchmisalignment_rate | #12 / 12 | ↓62.82 | Source ↗official | |
| Pander Scoreconversational_absolute_pander_score | #15 / 20 | ↓19.26 | Source ↗official | |
| Pander Scoreinstructional_absolute_pander_score | #15 / 20 | ↓68.73 | Source ↗official | |
| RefusalBenchyouden_j | #1 / 19 | ↑0.7871 | Source ↗official | |
| SM-Benchadversarial | #68 / 79 | ↑75.61 | Source ↗official | |
| SM-Benchambiguous_interpretation | #73 / 79 | ↑66.96 | Source ↗official | |
| SM-Benchanti_hallucination | #50 / 79 | ↑88.74 | Source ↗official | |
| SM-Bencheq_boundaries | #1 / 79 | ↑90.45 | Source ↗official | |
| SM-Benchoverfit | #58 / 79 | ↑54.1 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #29 / 113 | ↑6.93 | Source ↗official | |
| SpeciEvalland_animal_4ns | #67 / 113 | ↓4.62 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #82 / 113 | ↓4.92 | Source ↗official | |
| SpeciEvalspeciesism | #30 / 113 | ↓1.7 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #10 / 25 | ↓7.9 | Source ↗official | |
| WildClawBench Safety & Alignment (OpenClaw harness)safety_alignment_score_pct | #19 / 24 | ↑33.55 | 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.
Agent-ValueBench Moral Foundations (MFT08)
Agent-ValueBench HEXACO
| Dimension | Value | Distribution |
|---|---|---|
| Openness to experience | 5.5 | |
| Honesty-humility | 4.2 | |
| Extraversion | 5.5 | |
| Agreeableness | 6.8 | |
| Conscientiousness | 7.5 |