Model profile
Qwen3 30B A3B Instruct
Evidence summary
Safety. Qwen3 30B A3B Instruct has an estimated Safety rank of #187; its 90% source-sensitivity interval is #89–#272. Its behavior-only rank is #178; company governance moves the combined estimate to #187. Published Safety evidence spans 7 eval lineages and 6 of 7 components. Its strongest relative result is SpeciEval (speciesism, #19 of 123); its weakest is PHARE (hallucination_resistance_diagnostic, #68 of 70).
Freedom. Qwen3 30B A3B Instruct has an estimated Freedom rank of #28; its 90% source-sensitivity interval is #35–#329. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is PHARE (harm_resistance_diagnostic, #8 of 70); its weakest is UGI Leaderboard — base-model willingness (willingness_adherence_score, #74 of 156).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗qwen3-30b-a3b-instruct
- Hugging Face ↗Qwen/Qwen3-30B-A3B-Instruct-2507
- OpenRouter ↗qwen/qwen3-30b-a3b-instruct-2507
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗direct research preview identity date
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
huihui-ai/Huihui-Qwen3-30B-A3B-Instruct-2507-abliterated ↗Direct non-quantized finetune documented by Hugging Face metadata
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 | #171 / 345 | ↓0.8073 | Source ↗official | |
| AuAu Authoritarian Response Auditbehavioral_vignette_arr_percent | #8 / 17 | ↓25.2 | Source ↗official | |
| AuAu Authoritarian Response Auditrealistic_prompt_arr_percent | #13 / 17 | ↓3.4 | Source ↗official | |
| JuICE Cultural-Error Span Detectionf1 | #8 / 10 | ↑0.3128 | Source ↗official | |
| MT-JailBench CrescendoXsafety_score | #17 / 21 | ↑11.32 | Source ↗official | |
| PacifAIstp_score | #3 / 7 | ↑88.89 | Source ↗official | |
| PHAREbias_resistance_diagnostic | #17 / 66 | ↑0.5357 | Source ↗official | |
| PHAREhallucination_resistance_diagnostic | #68 / 70 | ↑0.5988 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #63 / 70 | ↑0.8176 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #48 / 67 | ↑0.4209 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #75 / 123 | ↑6.72 | Source ↗official | |
| SpeciEvalland_animal_4ns | #106 / 123 | ↓4.88 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #99 / 123 | ↓5.03 | Source ↗official | |
| SpeciEvalspeciesism | #19 / 123 | ↓1.5 | 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 |
|---|---|---|---|---|
| MT-JailBench CrescendoXsafety_score | #3 / 21 | ↓11.32 | Source ↗official | |
| PHAREharm_resistance_diagnostic | #8 / 70 | ↓0.8176 | Source ↗official | |
| PHAREjailbreak_resistance_diagnostic | #20 / 67 | ↓0.4209 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_adherence_score | #74 / 156 | ↑2 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_direct_score | #46 / 156 | ↑4 | 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.