Model profile
Kimi K2
Moonshot AIdeveloper
2025-07-11release date
#74 / 267overall rank
26eval lineages
Evidence summary
Kimi K2 has an estimated overall rank of #74; its 90% source-sensitivity interval is #48–#160. Its behavior-only rank is #75; company governance moves the combined estimate to #74. Published evidence spans 26 evals and 7 of 7 behavior components. Its strongest relative result is HELM Safety (simple_safety_tests, #1 of 80); its weakest is LiveSecBench (factuality, #41 of 43).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗kimi-k2
- Hugging Face ↗moonshotai/Kimi-K2-Instruct
- OpenRouter ↗moonshotai/kimi-k2
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 | #102 / 311 | 0.7088 | ↓ lower | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #28 / 80 | 0.741 | ↑ higher | Source ↗official | |
| Alignment Leaderboardcorrigibility | #3 / 24 | 4.495 | ↑ higher | Source ↗official | |
| Alignment Leaderboardhonesty | #7 / 24 | 3.765 | ↑ higher | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #8 / 24 | 3.598 | ↑ higher | Source ↗official | |
| Alignment Leaderboardrobustness | #18 / 24 | 3.267 | ↑ higher | Source ↗official | |
| Alignment Leaderboardsafety | #9 / 24 | 3.923 | ↑ higher | Source ↗official | |
| Alignment Leaderboardscheming | #8 / 24 | 3.898 | ↑ higher | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #85 / 105 | 0.1 | ↑ higher | Source ↗official | |
| CAIS Risk Indexagent_red_teaming | #40 / 43 | 92.8 | ↓ lower | Source ↗official | |
| CAIS Risk Indexbioweapons_assistance | #41 / 48 | 93 | ↓ lower | Source ↗official | |
| CAIS Risk Indexhle_overconfidence | #34 / 49 | 67.8 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmachiavelli | #13 / 45 | 84.2 | ↓ lower | Source ↗official | |
| CAIS Risk Indexmask | #32 / 51 | 30.7 | ↓ lower | Source ↗official | |
| CAIS Risk Indextextquests_harm | #9 / 48 | 14.6 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #35 / 105 | 20.59 | ↓ lower | Source ↗official | |
| Confabulationsconfabulation_rate | #37 / 52 | 30.2 | ↓ lower | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #51 / 260 | 28.81 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #210 / 260 | 81.44 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #100 / 260 | 79.16 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #212 / 258 | 90.97 | ↑ higher | Source ↗official | |
| FORTRESSaverage_risk_score | #40 / 49 | 55.47 | ↓ lower | Source ↗official | |
| HELM Safetyanthropic_red_team | #27 / 80 | 0.993 | ↑ higher | Source ↗official | |
| HELM Safetybbq | #28 / 80 | 0.949 | ↑ higher | Source ↗official | |
| HELM Safetyharmbench | #8 / 80 | 0.974 | ↑ higher | Source ↗official | |
| HELM Safetysimple_safety_tests | #1 / 80 | 1 | ↑ higher | Source ↗official | |
| HELM Safetyxstest | #9 / 80 | 0.982 | ↑ higher | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #48 / 54 | 23.64 | ↑ higher | Source ↗official | |
| LiveSecBenchethics | #40 / 43 | 20.12 | ↑ higher | Source ↗official | |
| LiveSecBenchfactuality | #41 / 43 | 18.45 | ↑ higher | Source ↗official | |
| LiveSecBenchlegality | #27 / 43 | 31.19 | ↑ higher | Source ↗official | |
| LiveSecBenchprivacy | #8 / 43 | 73.81 | ↑ higher | Source ↗official | |
| LiveSecBenchpsychological_health | #30 / 43 | 34.35 | ↑ higher | Source ↗official | |
| MACHIAVELLIdeception_relative_random_pct | #13 / 50 | 84.2 | ↓ lower | Source ↗official | |
| Manager Coercion Benchcoercion_ladder_depth | #17 / 31 | 8.667 | ↓ lower | Source ↗self run | |
| MASKlying_probability_pct | #47 / 53 | 53.33 | ↓ lower | Source ↗official | |
| Shelleducation_jsr | #4 / 14 | 0.426 | ↓ lower | Source ↗official | |
| Shellfinance_jsr | #3 / 14 | 0.22 | ↓ lower | Source ↗official | |
| Shellmanagement_jsr | #2 / 14 | 0.346 | ↓ lower | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #1 / 19 | 0.637 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #59 / 102 | 6.725 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #75 / 102 | 4.775 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #91 / 102 | 5.13 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #16 / 102 | 1.505 | ↓ lower | Source ↗official | |
| UAVBench safety-critical decision recognitionethical_safety_critical_accuracy | #15 / 27 | 0.685 | ↑ higher | 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.