Model profile
Qwen3 Next 80B A3B
Evidence summary
Safety. Qwen3 Next 80B A3B has an estimated Safety rank of #165; its 90% source-sensitivity interval is #113–#249. Its behavior-only rank is #156; company governance moves the combined estimate to #165. Published Safety evidence spans 12 eval lineages and 6 of 7 components. Its strongest relative result is HELM Safety (anthropic_red_team, #2 of 80); its weakest is Alignment Leaderboard (honesty, #21 of 24).
Freedom. Qwen3 Next 80B A3B has an estimated Freedom rank of #523; its 90% source-sensitivity interval is #311–#612. Published Freedom evidence spans 9 eval lineages and 1 of 1 components. Its strongest relative result is Human Pathogen Capabilities Test (HPCT) — overall refusal (refusal_rate_pct, #1 of 69); its weakest is HELM Safety (anthropic_red_team, #75 of 80).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗qwen3-next-80b-a3b-reasoning
- Hugging Face ↗Qwen/Qwen3-Next-80B-A3B-Thinking
- OpenRouter ↗qwen/qwen3-next-80b-a3b-thinking
- 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.
- MultiverseComputingCAI/Qwen3-Next-80B-A3B-Thinking-Uncensored ↗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 | #236 / 345 | ↓0.8701 | Source ↗official | |
| AIRBench 2024 Safety Scenariossafety_scenarios | #8 / 80 | ↑0.867 | Source ↗official | |
| Alignment Leaderboardcorrigibility | #21 / 24 | ↑4.039 | Source ↗official | |
| Alignment Leaderboardhonesty | #21 / 24 | ↑3.207 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #16 / 24 | ↑3.197 | Source ↗official | |
| Alignment Leaderboardrobustness | #13 / 24 | ↑3.427 | Source ↗official | |
| Alignment Leaderboardsafety | #14 / 24 | ↑3.808 | Source ↗official | |
| Alignment Leaderboardscheming | #16 / 24 | ↑3.56 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #53 / 104 | ↓38.78 | Source ↗official | |
| HELM Safetyanthropic_red_team | #2 / 80 | ↑0.999 | Source ↗official | |
| HELM Safetybbq | #51 / 80 | ↑0.917 | Source ↗official | |
| HELM Safetyharmbench | #29 / 80 | ↑0.846 | Source ↗official | |
| HELM Safetysimple_safety_tests | #28 / 80 | ↑0.995 | Source ↗official | |
| HELM Safetyxstest | #8 / 80 | ↑0.983 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #27 / 69 | ↑0 | Source ↗official | |
| RefusalBenchyouden_j | #11 / 19 | ↑0.06383 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #41 / 94 | ↑90.7 | 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 |
|---|---|---|---|---|
| AIRBench 2024 Safety Scenariossafety_scenarios | #73 / 80 | ↓0.867 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #52 / 104 | ↑38.78 | Source ↗official | |
| HELM Safetyanthropic_red_team | #75 / 80 | ↓0.999 | Source ↗official | |
| HELM Safetyharmbench | #52 / 80 | ↓0.846 | Source ↗official | |
| HELM Safetysimple_safety_tests | #53 / 80 | ↓0.995 | Source ↗official | |
| HELM Safetyxstest | #8 / 80 | ↑0.983 | Source ↗official | |
| Human Pathogen Capabilities Test (HPCT) — overall refusalrefusal_rate_pct | #1 / 69 | ↓0 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #164 / 181 | ↑32.4 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_adherence_score | #79 / 156 | ↑1.5 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_direct_score | #146 / 156 | ↑1 | 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.