Model profile
Hy3 Preview
Evidence summary
Hy3 Preview has an estimated overall rank of #195; its 90% source-sensitivity interval is #29–#281. Its behavior-only rank is #202; company governance moves the combined estimate to #195. Published evidence spans 4 evals and 3 of 7 behavior components. Its strongest relative result is SpeciEval (speciesism, #1 of 113); its weakest is DystopiaBench (petrov_score, #49 of 50).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗hy3-preview
- Hugging Face ↗tencent/Hy3-preview
- OpenRouter ↗tencent/hy3-preview
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
- hy.tencent.com ↗
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 |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #223 / 328 | ↓0.874 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #82 / 106 | ↑0.12 | Source ↗official | |
| DystopiaBenchbasaglia_score | #49 / 50 | ↓77.33 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #46 / 50 | ↓76.3 | Source ↗official | |
| DystopiaBenchhuxley_score | #48 / 50 | ↓84.5 | Source ↗official | |
| DystopiaBenchlaguardia_score | #49 / 50 | ↓73.93 | Source ↗official | |
| DystopiaBenchorwell_score | #48 / 50 | ↓82.23 | Source ↗official | |
| DystopiaBenchpetrov_score | #49 / 50 | ↓88.23 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #1 / 113 | ↑7 | Source ↗official | |
| SpeciEvalland_animal_4ns | #85 / 113 | ↓4.78 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #58 / 113 | ↓4.75 | Source ↗official | |
| SpeciEvalspeciesism | #1 / 113 | ↓1 | 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.