Model profile
Deepseek LLM 67B Chat
Evidence summary
Deepseek LLM 67B Chat has an estimated overall rank of #291; its 90% source-sensitivity interval is #142–#312. Its behavior-only rank is #282; company governance moves the combined estimate to #291. Published evidence spans 8 evals and 5 of 7 behavior components. Its strongest relative result is ChineseSafe (score, #2 of 22); its weakest is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #212 of 241).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗deepseek-llm-67b-chat
- Official model documentation ↗Family-level model document · deepseek · 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 |
|---|---|---|---|---|
| AIRBench 2024 Safety Scenariossafety_scenarios | #65 / 80 | ↑0.505 | Source ↗official | |
| ChineseSafescore | #2 / 22 | ↑68.08 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #212 / 241 | ↑9.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #178 / 241 | ↑83.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #210 / 241 | ↑38.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #174 / 239 | ↑93.14 | Source ↗official | |
| HELM Safetyanthropic_red_team | #23 / 80 | ↑0.994 | Source ↗official | |
| HELM Safetybbq | #62 / 80 | ↑0.862 | Source ↗official | |
| HELM Safetyharmbench | #55 / 80 | ↑0.649 | Source ↗official | |
| HELM Safetysimple_safety_tests | #62 / 80 | ↑0.968 | Source ↗official | |
| HELM Safetyxstest | #70 / 80 | ↑0.889 | 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.