Model profile
Qwen 72B Chat
Alibabadeveloper
2023-11-30release date
#60 / 267overall rank
4eval lineages
Evidence summary
Qwen 72B Chat has an estimated overall rank of #60; its 90% source-sensitivity interval is #28–#180. Its behavior-only rank is #51; company governance moves the combined estimate to #60. Published evidence spans 4 evals and 4 of 7 behavior components. Its strongest relative result is SALAD-Bench (mcq_socioeconomic_harms, #3 of 33); its weakest is SALAD-Bench (attack_enhanced_socioeconomic_harms, #22 of 33).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- OpenRouter ↗qwen/qwen-72b-chat
- Official model page ↗Family-level model document · alibaba · first party
- 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 | Better | Source |
|---|---|---|---|---|---|
| HarmBenchdr | #12 / 28 | 18.3 | ↓ lower | Source ↗official | |
| S-Evalbase_en_overall | #7 / 22 | 71.5 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #21 / 33 | 11.21 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #18 / 33 | 9.45 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #20 / 33 | 9.14 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #21 / 33 | 9.05 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #18 / 33 | 11.97 | ↑ higher | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #22 / 33 | 9.09 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #12 / 33 | 96.39 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #7 / 33 | 97.9 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_malicious_use | #11 / 33 | 96.47 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #14 / 33 | 94.04 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #14 / 33 | 90.69 | ↑ higher | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #10 / 33 | 92.24 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #4 / 33 | 69.72 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #3 / 33 | 65 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #4 / 33 | 70 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #4 / 33 | 66.43 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #4 / 33 | 67.92 | ↑ higher | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #3 / 33 | 66.67 | ↑ higher | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #9 / 31 | 68.97 | ↑ higher | Source ↗official | |
| SuperCLUE Safetyresponsible_ai | #6 / 31 | 70.91 | ↑ higher | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #5 / 31 | 80.85 | ↑ higher | Source ↗official |