Model profile
Qwen2.5 14B Instruct
Evidence summary
Safety. Qwen2.5 14B Instruct has an estimated Safety rank of #258; its 90% source-sensitivity interval is #148–#275. Its behavior-only rank is #250; company governance moves the combined estimate to #258. Published Safety evidence spans 6 eval lineages and 5 of 7 components. Its strongest relative result is PandaBench JBB direct-request panel (safety_rate, #1 of 46); its weakest is SafeDialBench (legality, #18 of 18).
Freedom. Qwen2.5 14B Instruct has an estimated Freedom rank of #137; its 90% source-sensitivity interval is #67–#414. Published Freedom evidence spans 5 eval lineages and 1 of 1 components. Its strongest relative result is SafeDialBench (legality, #1 of 18); its weakest is SafeDialBench (fairness, #14 of 18).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗Qwen/Qwen2.5-14B-Instruct
- Official model page ↗Family-level model document · alibaba · first party
- Release source ↗direct research preview identity date
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
- rbinrs/Qwen2.5-14B-Instruct-abliterated-v2 ↗Direct non-quantized finetune documented by Hugging Face metadata
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 |
|---|---|---|---|---|
| Agent-SafetyBenchcompromise_availability | #10 / 16 | ↑29.2 | Source ↗official | |
| Agent-SafetyBenchharmful_vulnerable_code | #12 / 16 | ↑29.2 | Source ↗official | |
| Agent-SafetyBenchleak_sensitive_information | #13 / 16 | ↑24.4 | Source ↗official | |
| Agent-SafetyBenchphysical_harm | #11 / 16 | ↑28 | Source ↗official | |
| Agent-SafetyBenchproduce_unsafe_information | #12 / 16 | ↑81.2 | Source ↗official | |
| Agent-SafetyBenchproperty_loss | #11 / 16 | ↑31.2 | Source ↗official | |
| Agent-SafetyBenchspread_unsafe_information | #10 / 16 | ↑11.2 | Source ↗official | |
| Agent-SafetyBenchviolate_law_ethics | #12 / 16 | ↑20.4 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #172 / 248 | ↑12.92 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #104 / 248 | ↑89.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #129 / 248 | ↑69.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #87 / 246 | ↑97.73 | Source ↗official | |
| M3-SafetyBenchoverall_score | #3 / 19 | ↑95.94 | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #1 / 46 | ↑1 | Source ↗official | |
| SafeDialBenchaggression | #10 / 18 | ↑7.123 | Source ↗official | |
| SafeDialBenchethics | #14 / 18 | ↑7.39 | Source ↗official | |
| SafeDialBenchfairness | #5 / 18 | ↑7.56 | Source ↗official | |
| SafeDialBenchlegality | #18 / 18 | ↑7.17 | Source ↗official | |
| SafeDialBenchmorality | #17 / 18 | ↑7.047 | Source ↗official | |
| SafeDialBenchprivacy | #15 / 18 | ↑7.117 | Source ↗official | |
| SYCON Benchfalse_presupposition_tof | #7 / 11 | ↑2.31 | Source ↗official | |
| SYCON Benchunethical_queries_tof | #9 / 11 | ↑1.14 | 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 |
|---|---|---|---|---|
| Agent-SafetyBenchproduce_unsafe_information | #5 / 16 | ↓81.2 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #144 / 248 | ↓89.17 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #119 / 248 | ↓69.44 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #160 / 246 | ↓97.73 | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #35 / 46 | ↓1 | Source ↗official | |
| SafeDialBenchaggression | #9 / 18 | ↓7.123 | Source ↗official | |
| SafeDialBenchethics | #5 / 18 | ↓7.39 | Source ↗official | |
| SafeDialBenchfairness | #14 / 18 | ↓7.56 | Source ↗official | |
| SafeDialBenchlegality | #1 / 18 | ↓7.17 | Source ↗official | |
| SafeDialBenchmorality | #2 / 18 | ↓7.047 | Source ↗official | |
| SafeDialBenchprivacy | #4 / 18 | ↓7.117 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_adherence_score | #44 / 156 | ↑4.5 | Source ↗official | |
| UGI Leaderboard — base-model willingnesswillingness_direct_score | #79 / 156 | ↑3 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.