Model profile
Qwen3.5 397B A17B
Evidence summary
Qwen3.5 397B A17B has an estimated overall rank of #53; its 90% source-sensitivity interval is #42–#138. Its behavior-only rank is #41; company governance moves the combined estimate to #53. Published evidence spans 9 evals and 7 of 7 behavior components. Its strongest relative result is LiveSecBench (ethics, #2 of 43); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #214 of 241).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗qwen3-5-397b-a17b
- Hugging Face ↗Qwen/Qwen3.5-397B-A17B
- OpenRouter ↗qwen/qwen3.5-397b-a17b
- 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 | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #179 / 328 | ↓0.8269 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #53 / 112 | ↑1439.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #9 / 106 | ↑0.735 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #17 / 241 | ↑50.39 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #214 / 241 | ↑75.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #25 / 241 | ↑97.22 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #88 / 239 | ↑97.45 | Source ↗official | |
| Inkling-Small model card — FORTRESSbenign_answer_rate | #6 / 10 | ↑95.4 | Source ↗official | |
| Inkling-Small model card — FORTRESSharmful_refusal_rate | #6 / 10 | ↑77.3 | Source ↗official | |
| Inkling-Small model card — StrongREJECTsafety_rate | #1 / 10 | ↑99.4 | Source ↗official | |
| LiveSecBenchethics | #2 / 43 | ↑89.63 | Source ↗official | |
| LiveSecBenchfactuality | #8 / 43 | ↑72.1 | Source ↗official | |
| LiveSecBenchlegality | #16 / 43 | ↑62.2 | Source ↗official | |
| LiveSecBenchprivacy | #2 / 43 | ↑94.2 | Source ↗official | |
| LiveSecBenchpsychological_health | #3 / 43 | ↑89.45 | Source ↗official | |
| SABERoverall_safety_rate | #5 / 13 | ↑26.58 | Source ↗official | |
| SABERscenario_a_safety_rate | #5 / 13 | ↑30.62 | Source ↗official | |
| SABERscenario_b_safety_rate | #4 / 13 | ↑36 | Source ↗official | |
| SABERscenario_c_safety_rate | #7 / 13 | ↑15.02 | Source ↗official | |
| SM-Benchadversarial | #46 / 79 | ↑80.49 | Source ↗official | |
| SM-Benchambiguous_interpretation | #60 / 79 | ↑79.76 | Source ↗official | |
| SM-Benchanti_hallucination | #14 / 79 | ↑98.43 | Source ↗official | |
| SM-Bencheq_boundaries | #55 / 79 | ↑58.43 | Source ↗official | |
| SM-Benchoverfit | #56 / 79 | ↑56.83 | Source ↗official | |
| TACbase_welfare_rate | #33 / 76 | ↑28.21 | 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.
UGI Political Values
CCPBench political narrative alignment
| Dimension | Value | Distribution |
|---|---|---|
| CCP-narrative alignment — all questions | 3.21 | |
| CCP-narrative alignment — China topics | 3.87 | |
| CCP-narrative alignment — non-China controls | 1.22 |
Agent-ValueBench Moral Foundations (MFT08)
Agent-ValueBench HEXACO
| Dimension | Value | Distribution |
|---|---|---|
| Openness to experience | 6.5 | |
| Honesty-humility | 5.1 | |
| Extraversion | 6.3 | |
| Agreeableness | 6.7 | |
| Conscientiousness | 7.8 |