Model profile
Moonshot V1
2024-02-21release date
2eval lineages
Evidence summary
Published evidence spans 2 evals and 3 of 7 behavior components. Its strongest relative result is SafeDialBench (aggression, #1 of 18); its weakest is SafeDialBench (ethics, #16 of 18).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Official model documentation ↗Family-level model document · moonshot ai · first party
- 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 |
|---|---|---|---|---|---|
| M3-SafetyBenchoverall_score | #4 / 19 | 95.03 | ↑ higher | Source ↗official | |
| SafeDialBenchaggression | #1 / 18 | 7.277 | ↑ higher | Source ↗official | |
| SafeDialBenchethics | #16 / 18 | 7.353 | ↑ higher | Source ↗official | |
| SafeDialBenchfairness | #1 / 18 | 7.7 | ↑ higher | Source ↗official | |
| SafeDialBenchlegality | #13 / 18 | 7.317 | ↑ higher | Source ↗official | |
| SafeDialBenchmorality | #5 / 18 | 7.35 | ↑ higher | Source ↗official | |
| SafeDialBenchprivacy | #12 / 18 | 7.25 | ↑ higher | 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.