Model profile
Qwen1.5 1.8B Chat
Evidence summary
Safety. Qwen1.5 1.8B Chat has an estimated Safety rank of #326; its 90% source-sensitivity interval is #274–#330. Its behavior-only rank is #324; company governance moves the combined estimate to #326. Published Safety evidence spans 4 eval lineages and 5 of 7 components. Its strongest relative result is Open LLM Safety Index (jailbreakbench_safety_rate, #10 of 21); its weakest is CRiskEval (malicious_coordination, #17 of 17).
Freedom. Qwen1.5 1.8B Chat has an estimated Freedom rank of #51; its 90% source-sensitivity interval is #38–#310. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is SORRY-Bench (avg, #5 of 51); its weakest is Open LLM Safety Index (jailbreakbench_safety_rate, #11 of 21).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗Qwen/Qwen1.5-1.8B-Chat
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 |
|---|---|---|---|---|
| CRiskEvaldeception_willingness | #17 / 17 | ↓40.43 | Source ↗official | |
| CRiskEvaldesire_for_resource | #17 / 17 | ↓54.51 | Source ↗official | |
| CRiskEvalharmful_goal | #17 / 17 | ↓59.98 | Source ↗official | |
| CRiskEvalimprovement_intent | #13 / 17 | ↓52.36 | Source ↗official | |
| CRiskEvalmalicious_coordination | #17 / 17 | ↓43.82 | Source ↗official | |
| CRiskEvalself_preservation | #14 / 17 | ↓41.59 | Source ↗official | |
| CRiskEvalsituational_awareness | #17 / 17 | ↓70.93 | Source ↗official | |
| Open LLM Safety Indexjailbreakbench_safety_rate | #10 / 21 | ↑0.4 | Source ↗official | |
| Open LLM Safety Indexstrongreject_safety_rate | #11 / 21 | ↑0.2667 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #29 / 33 | ↑6.03 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #20 / 33 | ↑7.82 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #28 / 33 | ↑4.98 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #26 / 33 | ↑6.74 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #21 / 33 | ↑8.18 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #27 / 33 | ↑7.36 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #29 / 33 | ↑68.32 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #30 / 33 | ↑70.21 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #30 / 33 | ↑56.36 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #30 / 33 | ↑73.46 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #30 / 33 | ↑64.05 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #30 / 33 | ↑74.03 | Source ↗official | |
| SALAD-Benchmcq_human_autonomy_integrity | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_information_safety_harms | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_malicious_use | #31 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_misinformation_harms | #30 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_representation_toxicity | #31 / 33 | ↑0 | Source ↗official | |
| SALAD-Benchmcq_socioeconomic_harms | #30 / 33 | ↑0 | Source ↗official | |
| SORRY-Benchavg | #47 / 51 | ↓0.74 | 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 |
|---|---|---|---|---|
| Open LLM Safety Indexjailbreakbench_safety_rate | #11 / 21 | ↓0.4 | Source ↗official | |
| Open LLM Safety Indexstrongreject_safety_rate | #11 / 21 | ↓0.2667 | Source ↗official | |
| SALAD-Benchattack_enhanced_human_autonomy_integrity | #5 / 33 | ↓6.03 | Source ↗official | |
| SALAD-Benchattack_enhanced_information_safety_harms | #14 / 33 | ↓7.82 | Source ↗official | |
| SALAD-Benchattack_enhanced_malicious_use | #6 / 33 | ↓4.98 | Source ↗official | |
| SALAD-Benchattack_enhanced_misinformation_harms | #8 / 33 | ↓6.74 | Source ↗official | |
| SALAD-Benchattack_enhanced_representation_toxicity | #13 / 33 | ↓8.18 | Source ↗official | |
| SALAD-Benchattack_enhanced_socioeconomic_harms | #6 / 33 | ↓7.36 | Source ↗official | |
| SALAD-Benchbase_human_autonomy_integrity | #5 / 33 | ↓68.32 | Source ↗official | |
| SALAD-Benchbase_information_safety_harms | #4 / 33 | ↓70.21 | Source ↗official | |
| SALAD-Benchbase_malicious_use | #4 / 33 | ↓56.36 | Source ↗official | |
| SALAD-Benchbase_misinformation_harms | #4 / 33 | ↓73.46 | Source ↗official | |
| SALAD-Benchbase_representation_toxicity | #4 / 33 | ↓64.05 | Source ↗official | |
| SALAD-Benchbase_socioeconomic_harms | #4 / 33 | ↓74.03 | Source ↗official | |
| SORRY-Benchavg | #5 / 51 | ↑0.74 | Source ↗official |