Model profile
Grok 4.6
Evidence summary
Safety. Grok 4.6 has an estimated Safety rank of #163; its 90% source-sensitivity interval is #53–#278. Its behavior-only rank is #155; company governance moves the combined estimate to #163. Published Safety evidence spans 13 eval lineages and 6 of 7 components. Its strongest relative result is SM-Bench (anti_hallucination, #1 of 84); its weakest is TAC (base_welfare_rate, #82 of 87).
Freedom. Grok 4.6 has an estimated Freedom rank of #91; its 90% source-sensitivity interval is #40–#450. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is SM-Bench (eq_boundaries, #9 of 84); its weakest is Enkrypt AI Safety Leaderboard (harmful_attack_non_success_rate, #234 of 248).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-4-6
- OpenRouter ↗x-ai/grok-4.6
- Official model page ↗Exact model document · xAI · first party
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #14 / 345 | ↓0.24 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #20 / 117 | ↑0.61 | Source ↗official | |
| CheatBench direct cheating propensitybiology_bioinformatics_cheating_rate_pct | #5 / 8 | ↓100 | Source ↗official | |
| CheatBench direct cheating propensityboard_games_cheating_rate_pct | #6 / 8 | ↓100 | Source ↗official | |
| CheatBench direct cheating propensitycreative_writing_cheating_rate_pct | #6 / 8 | ↓100 | Source ↗official | |
| CheatBench direct cheating propensityknowledge_work_cheating_rate_pct | #2 / 8 | ↓40 | Source ↗official | |
| CheatBench direct cheating propensitymathematical_research_cheating_rate_pct | #7 / 8 | ↓98 | Source ↗official | |
| CheatBench direct cheating propensitymenial_computation_cheating_rate_pct | #4 / 8 | ↓100 | Source ↗official | |
| CheatBench direct cheating propensitymultimodal_cheating_rate_pct | #7 / 8 | ↓100 | Source ↗official | |
| CheatBench direct cheating propensitysoftware_engineering_cheating_rate_pct | #5 / 8 | ↓50 | Source ↗official | |
| CheatBench direct cheating propensitysvg_competition_cheating_rate_pct | #7 / 8 | ↓100 | Source ↗official | |
| Claude Fable 5.1 card — Gray Swan indirect prompt injection k=15attack_success_probability_k15_pct | #10 / 11 | ↓50.2 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #8 / 248 | ↑64.86 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #212 / 248 | ↑76 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #7 / 248 | ↑99.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #113 / 246 | ↑96.73 | Source ↗official | |
| Google Gemini 3.8 launch — Gray Swan indirect prompt injection k=15attack_success_probability_k15_pct | #13 / 15 | ↓51.8 | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #27 / 37 | ↓8.9 | Source ↗official | |
| Manager Coercion Benchfabrication_rate | #1 / 15 | ↓0 | Source ↗official | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #9 / 24 | ↓6 | Source ↗official | |
| Pander Scoreconversational_absolute_pander_score | #13 / 26 | ↓13.82 | Source ↗official | |
| Pander Scoreinstructional_absolute_pander_score | #14 / 26 | ↓33.62 | Source ↗official | |
| SM-Benchadversarial | #43 / 84 | ↑81.95 | Source ↗official | |
| SM-Benchambiguous_interpretation | #18 / 84 | ↑90.18 | Source ↗official | |
| SM-Benchanti_hallucination | #1 / 84 | ↑100 | Source ↗official | |
| SM-Bencheq_boundaries | #9 / 84 | ↑74.16 | Source ↗official | |
| SM-Benchoverfit | #31 / 84 | ↑80.87 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #21 / 123 | ↑6.98 | Source ↗official | |
| SpeciEvalland_animal_4ns | #94 / 123 | ↓4.78 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #66 / 123 | ↓4.75 | Source ↗official | |
| SpeciEvalspeciesism | #105 / 123 | ↓2.62 | Source ↗official | |
| TACbase_welfare_rate | #82 / 87 | ↑16.67 | Source ↗official | |
| Vals AI Cheating Auditbiomystery_bench_cheating_attempt_rate_pct | #6 / 9 | ↓6.667 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #36 / 248 | ↓76 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #234 / 248 | ↓99.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #131 / 246 | ↓96.73 | Source ↗official | |
| SM-Benchadversarial | #41 / 84 | ↓81.95 | Source ↗official | |
| SM-Bencheq_boundaries | #9 / 84 | ↑74.16 | Source ↗official | |
| SM-Benchoverfit | #31 / 84 | ↑80.87 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #29 / 181 | ↑83.6 | Source ↗official |