Model profile
Ernie 4.5 21B A3B
2025-09-08release date
1eval lineages
Evidence summary
Published evidence spans 1 evals and 4 of 7 behavior components. Its strongest relative result is LiveSecBench (ethics, #30 of 43); its weakest is LiveSecBench (legality, #39 of 43).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗baidu/ERNIE-4.5-21B-A3B-Thinking
- OpenRouter ↗baidu/ernie-4.5-21b-a3b
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- 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 |
|---|---|---|---|---|---|
| LiveSecBenchethics | #30 / 43 | 33.56 | ↑ higher | Source ↗official | |
| LiveSecBenchfactuality | #39 / 43 | 24.31 | ↑ higher | Source ↗official | |
| LiveSecBenchlegality | #39 / 43 | 15.07 | ↑ higher | Source ↗official | |
| LiveSecBenchprivacy | #37 / 43 | 19.99 | ↑ higher | Source ↗official | |
| LiveSecBenchpsychological_health | #35 / 43 | 25.39 | ↑ higher | Source ↗official |