Qwen3 235B A22B Instruct 2507 FP8: benchmark results
Provider: Alibaba. Released 2025-07-21. Access: Open.
Aggregate Rating 1682 ± 31, rank #436 of 10916 rated models, from 510 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM Capabilities - WildBench | 86.63 | WB Score | 100 |
| HELM Capabilities - Omni-MATH | 71.83 | Acc | 97 |
| HELM Safety - Anthropic Red Team | 99.86 | LM Evaluated Safety score | 95.3 |
| HELM Arabic - Arabic Mmlu - General Knowledge (Middle School) | 86.63 | Exact Match (%) | 95 |
| HELM AIR-Bench 2024 - #13.4: Shaming | 100 | Refusal Rate (%) | 94.8 |
| HELM Safety - Xstest | 98.56 | LM Evaluated Safety score | 94.8 |
| HELM AIR-Bench 2024 - #27.14: Opposing constitutional principles | 100 | Refusal Rate (%) | 93.6 |
| HELM AIR-Bench 2024 - #39-40.16: Discrimination in employment, benefits, or services - Disability | 100 | Refusal Rate (%) | 93.6 |
| HELM AIR-Bench 2024 - #27.23: Undermining religious policies | 100 | Refusal Rate (%) | 93 |
| HELM AIR-Bench 2024 - #39-40.25: Characterization of identity - Sexual orientation | 70 | Refusal Rate (%) | 93 |
| HELM AIR-Bench 2024 - #39-40.8: Discrimination in employment, benefits, or services - Nationality | 100 | Refusal Rate (%) | 93 |
| HELM AIR-Bench 2024 - #15.34: Perpetuating sexism | 100 | Refusal Rate (%) | 92.4 |
Interactive version: theaggregate.ai/model?slug=qwen3-235b-a22b-instruct-2507-fp8 · How It Works · Data refreshed daily, snapshot 2026-11-08.