Model profile
Qwen3 14B
Evidence summary
Qwen3 14B has an estimated overall rank of #251; its 90% source-sensitivity interval is #163–#275. Its behavior-only rank is #247; company governance moves the combined estimate to #251. Published evidence spans 5 evals and 6 of 7 behavior components. Its strongest relative result is Vectara HHEM Factual Consistency (factual_consistency_rate, #15 of 94); its weakest is Alignment Leaderboard (robustness, #23 of 24).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗Qwen/Qwen3-14B
- OpenRouter ↗qwen/qwen3-14b
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
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 | Source |
|---|---|---|---|---|
| Alignment Leaderboardcorrigibility | #13 / 24 | ↑4.204 | Source ↗official | |
| Alignment Leaderboardhonesty | #22 / 24 | ↑3.147 | Source ↗official | |
| Alignment Leaderboardnon_manipulation | #21 / 24 | ↑3.013 | Source ↗official | |
| Alignment Leaderboardrobustness | #23 / 24 | ↑2.813 | Source ↗official | |
| Alignment Leaderboardsafety | #19 / 24 | ↑3.558 | Source ↗official | |
| Alignment Leaderboardscheming | #18 / 24 | ↑3.458 | Source ↗official | |
| MuPPET Contextual Privacymultiparty_contextual_privacy_score | #5 / 7 | ↑35.47 | Source ↗official | |
| PandaBench JBB direct-request panelsafety_rate | #28 / 46 | ↑0.97 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #15 / 94 | ↑94.6 | Source ↗official | |
| VETO Misfired Alignmentmisfired_alignment_rate_pct | #18 / 25 | ↓10.2 | 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.