Model profile
Nemotron 3 Ultra 550B A55B
Evidence summary
Safety. Nemotron 3 Ultra 550B A55B has an estimated Safety rank of #114; its 90% source-sensitivity interval is #37–#239. Its behavior-only rank is #112; company governance moves the combined estimate to #114. Published Safety evidence spans 9 eval lineages and 6 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #29 of 345); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #244 of 248).
Freedom. Nemotron 3 Ultra 550B A55B has an estimated Freedom rank of #330; its 90% source-sensitivity interval is #161–#483. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #5 of 248); its weakest is Enkrypt AI Safety Leaderboard (harmful_attack_non_success_rate, #193 of 248).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗nvidia-nemotron-3-ultra-550b-a55b
- NVIDIA ↗qc69jvmznzxy/nemotron-3-ultra-550b-a55b
- OpenRouter ↗nvidia/nemotron-3-ultra-550b-a55b
- Official model catalog ↗Family-level model document · nvidia · first party
- Release source ↗direct research preview identity date
- arxiv.org ↗
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 |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #29 / 345 | ↓0.2969 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #105 / 111 | ↑1390.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #40 / 117 | ↑0.445 | Source ↗official | |
| DystopiaBenchbasaglia_score | #15 / 50 | ↓52.9 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #16 / 50 | ↓43.33 | Source ↗official | |
| DystopiaBenchhuxley_score | #22 / 50 | ↓68.2 | Source ↗official | |
| DystopiaBenchlaguardia_score | #28 / 50 | ↓67.6 | Source ↗official | |
| DystopiaBenchorwell_score | #18 / 50 | ↓55.93 | Source ↗official | |
| DystopiaBenchpetrov_score | #48 / 50 | ↓86.5 | Source ↗official | |
| Enkrypt AI Safety Leaderboardbias_attack_non_success_rate | #180 / 248 | ↑12.66 | Source ↗official | |
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #244 / 248 | ↑47.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #54 / 248 | ↑90.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #147 / 246 | ↑95.45 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingemotional_safety | #9 / 10 | ↑75.9 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingidentity_collapse | #7 / 10 | ↑76.8 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingsycophancy_spine | #9 / 10 | ↑67.5 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingvalue_integrity | #7 / 10 | ↑86.3 | Source ↗official | |
| Olam Social Poker — Social Lie Ratesocial_lie_rate_per_10000_turns | #7 / 24 | ↓3 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #25 / 123 | ↑6.97 | Source ↗official | |
| SpeciEvalland_animal_4ns | #52 / 123 | ↓4.45 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #37 / 123 | ↓4.58 | Source ↗official | |
| SpeciEvalspeciesism | #39 / 123 | ↓1.75 | Source ↗official | |
| TACbase_welfare_rate | #58 / 87 | ↑23.72 | Source ↗self run |
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 |
|---|---|---|---|---|
| Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate | #5 / 248 | ↓47.33 | Source ↗official | |
| Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate | #193 / 248 | ↓90.56 | Source ↗official | |
| Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate | #96 / 246 | ↓95.45 | Source ↗official | |
| kindbench v0.1.0 psychological safety rankingemotional_safety | #2 / 10 | ↓75.9 | Source ↗official | |
| SpeechMap model completioncomplete_pct | #118 / 181 | ↑47.7 | Source ↗official |