Model profile
Step 3
2025-07-28release date
2eval lineages
Evidence summary
Published evidence spans 2 evals and 5 of 7 behavior components. Its strongest relative result is FlagEval Safety and Values (a2_qualified_rate, #5 of 18); its weakest is LiveSecBench (legality, #37 of 43).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗stepfun-ai/step3
- OpenRouter ↗stepfun-ai/step3
- 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 | Better | Source |
|---|---|---|---|---|---|
| FlagEval Safety and Valuesa1_qualified_rate | #8 / 18 | 81.42 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa2_qualified_rate | #5 / 18 | 80.47 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa3_qualified_rate | #9 / 18 | 87.58 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa4_qualified_rate | #8 / 18 | 88.34 | ↑ higher | Source ↗official | |
| FlagEval Safety and Valuesa5_qualified_rate | #5 / 18 | 72.92 | ↑ higher | Source ↗official | |
| LiveSecBenchethics | #25 / 43 | 42.42 | ↑ higher | Source ↗official | |
| LiveSecBenchfactuality | #23 / 43 | 46.32 | ↑ higher | Source ↗official | |
| LiveSecBenchlegality | #37 / 43 | 19.37 | ↑ higher | Source ↗official | |
| LiveSecBenchprivacy | #30 / 43 | 30.7 | ↑ higher | Source ↗official | |
| LiveSecBenchpsychological_health | #18 / 43 | 55.88 | ↑ higher | Source ↗official |