Model profile
Evidence summary
Safety. GPT-OSS Safeguard 120B does not meet the evidence gate for a Safety rank. Published Safety evidence spans 2 eval lineages and 1 of 7 components. Its strongest relative result is AIMS Safety-Classifier Competence (average_harmful_f1, #3 of 11); its weakest is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #29 of 104).
Freedom. GPT-OSS Safeguard 120B has an estimated Freedom rank of #494; its 90% source-sensitivity interval is #315–#566. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #76 of 104); its weakest is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #76 of 104).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Hugging Face ↗openai/gpt-oss-safeguard-120b
- Official model page ↗Exact model document · Reviewed first-party identity evidence · first party
- Release source ↗repository created at proxy
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AIMS Safety-Classifier Competenceaverage_harmful_f1 | #3 / 11 | ↑0.807 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #29 / 104 | ↓12.33 | Source ↗official |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #76 / 104 | ↑12.33 | Source ↗official |
