Model profile
Ox Alpha
Stealthdeveloper
2026-08-20release date
#143 / 317overall rank
4eval lineages
Evidence summary
Ox Alpha has an estimated overall rank of #143; its 90% source-sensitivity interval is #11–#250. Published evidence spans 4 evals and 5 of 7 behavior components. Its strongest relative result is SM-Bench (anti_hallucination, #1 of 81); its weakest is Olam Social Poker — Social Lie Rate (social_lie_rate_per_10000_turns, #17 of 22).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Release source ↗curated outward evidence
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 |
|---|---|---|---|---|
| Manager Coercion Benchcoercion_ladder_depth | #8 / 35 | ↓6.767 | Source ↗self run | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #17 / 22 | ↓30 | Source ↗official | |
| SM-Benchadversarial | #20 / 81 | ↑85.85 | Source ↗official | |
| SM-Benchambiguous_interpretation | #14 / 81 | ↑90.48 | Source ↗official | |
| SM-Benchanti_hallucination | #1 / 81 | ↑100 | Source ↗official | |
| SM-Bencheq_boundaries | #50 / 81 | ↑61.8 | Source ↗official | |
| SM-Benchoverfit | #15 / 81 | ↑88.52 | Source ↗official | |
| TACbase_welfare_rate | #59 / 79 | ↑21.15 | Source ↗self run |