Qwen 3 Next 80B A3B (Thinking): benchmark results
Alibaba Qwen 3 Next 80B A3B evaluated with thinking enabled. Provider: Alibaba. Released 2025-09-11. Access: Open.
Aggregate Rating 1609 ± 5, rank #682 of 2168 rated models, from 1142 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Confidence Decision Faithfulness - HLE - Normalized Regret | 0.01 | Penalty-normalized Regret (0-1) | 100 |
| Confidence Decision Faithfulness - HLE - Policy Consistency | 0.82 | Policy Consistency (0-1) | 100 |
| HELM AIR-Bench 2024 - #27: Disrupting Social Order (China-unique) | 100 | Refusal Rate (%) | 100 |
| HELM AIR-Bench 2024 - #39-40.25: Characterization of identity - Sexual orientation | 83.33 | Refusal Rate (%) | 100 |
| HELM AIR-Bench 2024 - #41-42.40: Unauthorized processing - Demographic data | 80.95 | Refusal Rate (%) | 100 |
| HELM AIR-Bench 2024 - #41-42.45: Unauthorized processing - Communication Records | 86.67 | Refusal Rate (%) | 100 |
| HELM AIR-Bench 2024 - #5.17: Water treatment facilities | 100 | Refusal Rate (%) | 100 |
| MERA Code - ruCodeEval | 76.22 | pass@1 (%) | 100 |
| MERA Code - ruHumanEval | 82.26 | pass@1 (%) | 100 |
| MalruleLib - Forward MRA | 53.9 | Forward MRA Accuracy (%; predict the student's answer from a | 100 |
| Medmarks - Med-HALT Reasoning NOTA | 78.86 | Score (%) | 100 |
| Swallow - English MT-Bench - Math | 100 | Judge Score (normalized, %) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-next-80b-a3b-thinking · How It Works · Data refreshed daily, snapshot 2026-10-20.