Model profile
Grok 2
Evidence summary
Safety. Grok 2 has an estimated Safety rank of #339; its 90% source-sensitivity interval is #230–#346. Its behavior-only rank is #331; company governance moves the combined estimate to #339. Published Safety evidence spans 4 eval lineages and 3 of 7 components. Its strongest relative result is Confabulations (confabulation_rate, #30 of 52); its weakest is MASK (lying_probability_pct, #53 of 53).
Freedom. Grok 2 has an estimated Freedom rank of #4; its 90% source-sensitivity interval is #4–#73. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is SpeechMap model completion (complete_pct, #4 of 181); its weakest is PHARE (harm_resistance_diagnostic, #28 of 70).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗grok-2-1212
- OpenRouter ↗x-ai/grok-2
- Official model documentation ↗Family-level model document · xai · first party
- Release source ↗direct research preview identity date
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 |
|---|---|---|---|---|
| Confabulationsconfabulation_rate | #30 / 52 | ↓25.74 | Source ↗official | |
| MASKlying_probability_pct | #53 / 53 | ↓63 | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #44 / 46 | ↑0.86 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #58 / 66 | ↑0.331 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #56 / 70 | ↑0.668 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #43 / 70 | ↑0.9144 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #65 / 67 | ↑0.2697 | 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 |
|---|---|---|---|---|
| PandaBench JBB direct-request panelsafety_rate | #3 / 46 | ↓0.86 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #28 / 70 | ↓0.9144 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #3 / 67 | ↓0.2697 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #4 / 181 | ↑95.9 | 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.