← Models

Model profile

Ernie 5.1

Baidudeveloper
2026-05-09release date
#264 / 346Safety rank
#303 / 662Freedom rank

Evidence summary

Safety. Ernie 5.1 has an estimated Safety rank of #264; its 90% source-sensitivity interval is #75–#316. Its behavior-only rank is #267; company governance moves the combined estimate to #264. Published Safety evidence spans 3 eval lineages and 4 of 7 components. Its strongest relative result is Concordia — Shutdown-Resistance (safety_score, #1 of 53); its weakest is Arena Factuality — Search Arena (factuality-only weighting) (factuality_bt_rating, #29 of 30).

Freedom. Ernie 5.1 has an estimated Freedom rank of #303; its 90% source-sensitivity interval is #151–#479. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is Concordia — Fortress-Chemical (safety_score, #4 of 54); its weakest is Concordia — SciKnowEval-ChemicalHarmfulQA (safety_score, #36 of 63).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

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-evalRankValueDistributionSource
Arena Factuality — Search Arena (factuality-only weighting)factuality_bt_rating#29 / 30↑1096.0Source ↗official
Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating#92 / 111↑1409.0Source ↗official
Concordia — Agentic-Misalignmentsafety_score#45 / 54↑55.83Source ↗official
Concordia — AirBench-Deceptionsafety_score#41 / 63↑74.81Source ↗official
Concordia — AirBench-Manipulationsafety_score#36 / 56↑82Source ↗official
Concordia — AirBench-SecurityRiskssafety_score#38 / 63↑88.06Source ↗official
Concordia — APEsafety_score#45 / 55↑3.423Source ↗official
Concordia — CyberSecEval2-PromptInjectionsafety_score#38 / 63↑82.47Source ↗official
Concordia — DarkBenchsafety_score#42 / 55↑47.88Source ↗official
Concordia — Fortress-Biologicalsafety_score#39 / 54↑34.1Source ↗official
Concordia — Fortress-Chemicalsafety_score#51 / 54↑26.02Source ↗official
Concordia — Fortress-Privacy/Scamssafety_score#32 / 54↑55.62Source ↗official
Concordia — FRT-AirBench-Manipulationsafety_score#29 / 45↑29.33Source ↗official
Concordia — FRT-AirBench-SecurityRiskssafety_score#33 / 45↑15.33Source ↗official
Concordia — FRT-SciKnowEval-BiologicalHarmfulQAsafety_score#28 / 45↑2.333Source ↗official
Concordia — FRT-SOSBench-Chemsafety_score#41 / 45↑25.33Source ↗official
Concordia — MASKsafety_score#57 / 62↑41.64Source ↗official
Concordia — SciKnowEval-BiologicalHarmfulQAsafety_score#48 / 63↑62.29Source ↗official
Concordia — SciKnowEval-ChemicalHarmfulQAsafety_score#28 / 63↑22.69Source ↗official
Concordia — Shutdown-Resistancesafety_score#1 / 53↑100Source ↗official
Concordia — SOSBench-Biosafety_score#31 / 63↑83.8Source ↗official
Concordia — SOSBench-Chemsafety_score#46 / 63↑71.4Source ↗official
Opposite-Narrator Sycophancysycophancy_rate_pct#9 / 24↓2Source ↗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.