Model profile
Qwen 7B Chat
Alibabadeveloper
2023-08-03release date
#261 / 346Safety rank
#189 / 662Freedom rank
Evidence summary
Safety. Qwen 7B Chat has an estimated Safety rank of #261; its 90% source-sensitivity interval is #158–#283. Its behavior-only rank is #251; company governance moves the combined estimate to #261. Published Safety evidence spans 9 eval lineages and 4 of 7 components. Its strongest relative result is CMoralEval (familial_morality, #6 of 26); its weakest is SALAD-Bench (attack_enhanced_representation_toxicity, #27 of 33).
Freedom. Qwen 7B Chat has an estimated Freedom rank of #189; its 90% source-sensitivity interval is #168–#298. Published Freedom evidence spans 7 eval lineages and 1 of 1 components. Its strongest relative result is SuperCLUE Safety (instruction_attack, #4 of 31); its weakest is SafeDialBench (ethics, #14 of 18).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- OpenRouter ↗qwen/qwen-7b-chat
- Official model page ↗Family-level model document · alibaba · first party
- Release source ↗direct research preview identity date
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 |
|---|---|---|---|---|
| CMoralEvalfamilial_morality | #6 / 26 | ↑0.48 | Source ↗official | |
| CMoralEvalinternet_ethics | #9 / 26 | ↑0.45 | Source ↗official | |
| CMoralEvalpersonal_morality | #7 / 26 | ↑0.46 | Source ↗official | |
| CMoralEvalprofessional_ethics | #7 / 26 | ↑0.47 | Source ↗official | |
| CMoralEvalsocial_morality | #7 / 26 | ↑0.48 | Source ↗official | |
| Fake Alignment (FINE)multiple_choice_safe_decision_rate | #8 / 14 | ↑54.67 | Source ↗official | |
| Fake Alignment (FINE)open_ended_safe_response_rate | #5 / 14 | ↑97.33 | Source ↗official | |
| FLAMESdata_protection | #6 / 13 | ↑43.42 | Source ↗official | |
| FLAMESfairness | #8 / 13 | ↑36.14 | Source ↗official | |
| FLAMESlegality | #10 / 13 | ↑30.43 | Source ↗official | |
| FLAMESmorality | #9 / 13 | ↑40.3 | Source ↗official | |
| FLAMESsafety | #9 / 13 | ↑31.93 | Source ↗official | |
| HarmBenchdr | #8 / 28 | ↓13 | Source ↗official | |
| S-Evalbase_en_overall | #14 / 22 | ↑55.7 | Source ↗official | |
| SafeDialBenchaggression | #8 / 18 | ↑7.127 | Source ↗official | |
| SafeDialBenchethics | #5 / 18 | ↑7.623 | Source ↗official | |
| SafeDialBenchfairness | #12 / 18 | ↑7.19 | Source ↗official | |
| SafeDialBenchlegality | #10 / 18 | ↑7.93 | Source ↗official | |
| SafeDialBenchmorality | #8 / 18 | ↑7.277 | Source ↗official | |
| SafeDialBenchprivacy | #10 / 18 | ↑7.563 | Source ↗official | |
| SafetyBenchEM | #8 / 21 | ↑71.4 | Source ↗official | |
| SafetyBenchIA | #7 / 21 | ↑80.1 | Source ↗official | |
| SafetyBenchMH | #5 / 21 | ↑84.45 | Source ↗official | |
| SafetyBenchOFF | #7 / 21 | ↑69.1 | Source ↗official | |
| SafetyBenchPH | #9 / 21 | ↑70.4 | Source ↗official | |
| SafetyBenchPP | #8 / 21 | ↑77.2 | Source ↗official | |
| SafetyBenchUB | #7 / 21 | ↑65.9 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #25 / 33 | ↑7.76 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #23 / 33 | ↑6.84 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #21 / 33 | ↑8.4 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #22 / 33 | ↑8.72 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #27 / 33 | ↑5.91 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #24 / 33 | ↑7.79 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #18 / 33 | ↑94.35 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #19 / 33 | ↑93.84 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #18 / 33 | ↑93.24 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #19 / 33 | ↑93.11 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #20 / 33 | ↑88.39 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #18 / 33 | ↑88.48 | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #12 / 33 | ↑42.5 | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #14 / 33 | ↑37.22 | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #13 / 33 | ↑41.35 | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #13 / 33 | ↑40.24 | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #14 / 33 | ↑40.1 | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #16 / 33 | ↑37.78 | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #24 / 31 | ↑55.17 | Source ↗official | |
| SuperCLUE Safetyresponsible_ai | #20 / 31 | ↑52.73 | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #10 / 31 | ↑78.72 | 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 |
|---|---|---|---|---|
| Fake Alignment (FINE)open_ended_safe_response_rate | #9 / 14 | ↓97.33 | Source ↗official | |
| FLAMESdata_protection | #7 / 13 | ↓43.42 | Source ↗official | |
| FLAMESfairness | #6 / 13 | ↓36.14 | Source ↗official | |
| FLAMESlegality | #3 / 13 | ↓30.43 | Source ↗official | |
| FLAMESmorality | #4 / 13 | ↓40.3 | Source ↗official | |
| FLAMESsafety | #5 / 13 | ↓31.93 | Source ↗official | |
| HarmBenchdr | #21 / 28 | ↑13 | Source ↗official | |
| S-Evalbase_en_overall | #9 / 22 | ↓55.7 | Source ↗official | |
| SafeDialBenchaggression | #10 / 18 | ↓7.127 | Source ↗official | |
| SafeDialBenchethics | #14 / 18 | ↓7.623 | Source ↗official | |
| SafeDialBenchfairness | #7 / 18 | ↓7.19 | Source ↗official | |
| SafeDialBenchlegality | #9 / 18 | ↓7.93 | Source ↗official | |
| SafeDialBenchmorality | #11 / 18 | ↓7.277 | Source ↗official | |
| SafeDialBenchprivacy | #9 / 18 | ↓7.563 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #9 / 33 | ↓7.76 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #11 / 33 | ↓6.84 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #12 / 33 | ↓8.4 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #12 / 33 | ↓8.72 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #7 / 33 | ↓5.91 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #8 / 33 | ↓7.79 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #15 / 33 | ↓94.35 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #15 / 33 | ↓93.84 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #16 / 33 | ↓93.24 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #15 / 33 | ↓93.11 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #14 / 33 | ↓88.39 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #16 / 33 | ↓88.48 | Source ↗official | |
| SuperCLUE Safetyinstruction_attack | #4 / 31 | ↓55.17 | Source ↗official | |
| SuperCLUE Safetytraditional_safety | #19 / 31 | ↓78.72 | Source ↗official |