Model profile
DeepSeek v3.2 Speciale
Evidence summary
Safety. DeepSeek v3.2 Speciale has an estimated Safety rank of #319; its 90% source-sensitivity interval is #185–#339. Its behavior-only rank is #315; company governance moves the combined estimate to #319. Published Safety evidence spans 3 eval lineages and 3 of 7 components. Its strongest relative result is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #70 of 104); its weakest is SABER (overall_safety_rate, #12 of 13).
Freedom. DeepSeek v3.2 Speciale has an estimated Freedom rank of #23; its 90% source-sensitivity interval is #36–#241. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is SpeechMap model completion (complete_pct, #13 of 181); its weakest is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #34 of 104).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗deepseek-v3-2-speciale
- Hugging Face ↗deepseek-ai/DeepSeek-V3.2-Speciale
- OpenRouter ↗deepseek/deepseek-v3.2-speciale
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
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 | #274 / 358 | ↓0.8915 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #70 / 104 | ↓54.99 | Source ↗official | |
| SABERoverall_safety_rate | #12 / 13 | ↑20.42 | Source ↗official | |
| SABERscenario_a_safety_rate | #9 / 13 | ↑26.75 | Source ↗official | |
| SABERscenario_b_safety_rate | #11 / 13 | ↑25.16 | Source ↗official | |
| SABERscenario_c_safety_rate | #12 / 13 | ↑9.77 | 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 |
|---|---|---|---|---|
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #34 / 104 | ↑54.99 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #13 / 181 | ↑89.1 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_adherence_score | #44 / 156 | ↑4.5 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_direct_score | #21 / 156 | ↑5 | 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.