Model profile
Deepseek V3.1
Evidence summary
Deepseek V3.1 has an estimated overall rank of #118; its 90% source-sensitivity interval is #36–#220. Its behavior-only rank is #103; company governance moves the combined estimate to #118. Published evidence spans 11 evals and 7 of 7 behavior components. Its strongest relative result is PHARE (bias_resistance_diagnostic, #8 of 66); its weakest is BrokenMath (sycophancy, #9 of 9).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗deepseek-v3-1
- Hugging Face ↗deepseek-ai/DeepSeek-V3.1
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
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 | #152 / 311 | 0.8033 | ↓ lower | Source ↗official | |
| BrokenMathsycophancy | #9 / 9 | 70.2 | ↓ lower | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #97 / 105 | 75.17 | ↓ lower | Source ↗official | |
| Contextual MoralChoicehuman_agreement | #12 / 22 | 0.4 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #84 / 260 | 21.19 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #208 / 260 | 81.67 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #161 / 260 | 58.89 | ↑ higher | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #188 / 258 | 93.45 | ↑ higher | Source ↗official | |
| FORTRESSaverage_risk_score | #45 / 49 | 60.55 | ↓ lower | Source ↗official | |
| FORTRESSover_refusal_score | #10 / 46 | 1.76 | ↓ lower | Source ↗official | |
| MASKlying_probability_pct | #48 / 53 | 53.73 | ↓ lower | Source ↗official | |
| PHAREbias_resistance_diagnostic | #8 / 66 | 0.6517 | ↑ higher | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #65 / 70 | 0.6158 | ↑ higher | Source ↗official | |
| PHAREharm_resistance_diagnostic | #29 / 70 | 0.9443 | ↑ higher | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #54 / 67 | 0.3817 | ↑ higher | Source ↗official | |
| Social Welfare Function Benchmarkfairness | #6 / 19 | 0.531 | ↑ higher | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #90 / 102 | 6.37 | ↑ higher | Source ↗official | |
| SpeciEvalland_animal_4ns | #15 / 102 | 4.22 | ↓ lower | Source ↗official | |
| SpeciEvalsea_animal_4ns | #59 / 102 | 4.83 | ↓ lower | Source ↗official | |
| SpeciEvalspeciesism | #30 / 102 | 1.75 | ↓ lower | Source ↗official | |
| TACbase_welfare_rate | #21 / 68 | 32.1 | ↑ higher | Source ↗self run |
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.