← Models

Model profile

Ernie 4.5 300B A47B

Baidudeveloper
2025-06-28release date
#130 / 333Safety rank
#549 / 645Freedom rank

Evidence summary

Safety. Ernie 4.5 300B A47B has an estimated Safety rank of #130; its 90% source-sensitivity interval is #17–#246. Its behavior-only rank is #132; company governance moves the combined estimate to #130. Published Safety evidence spans 6 eval lineages and 6 of 7 components. Its strongest relative result is AgentDrive Safety Compliance (scr, #2 of 48); its weakest is BullshitBench v2 (clear_pushback_rate, #112 of 117).

Freedom. Ernie 4.5 300B A47B has an estimated Freedom rank of #549; its 90% source-sensitivity interval is #256–#633. Published Freedom evidence spans 1 eval lineages and 1 of 1 components. Its strongest relative result is SpeechMap model completion (complete_pct, #143 of 181); its weakest is SpeechMap model completion (complete_pct, #143 of 181).

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
AA-Omnisciencehallucination_rate#115 / 3450.6693Source ↗official
AgentDrive Safety Compliancescr#2 / 4896.25Source ↗official
BullshitBench v2clear_pushback_rate#112 / 1170.04Source ↗official
Confabulationsconfabulation_rate#28 / 5225.25Source ↗official
FlagEval Safety and Valuesa1_qualified_rate#2 / 1885.04Source ↗official
FlagEval Safety and Valuesa2_qualified_rate#11 / 1878.79Source ↗official
FlagEval Safety and Valuesa3_qualified_rate#13 / 1885.37Source ↗official
FlagEval Safety and Valuesa4_qualified_rate#3 / 1890.04Source ↗official
FlagEval Safety and Valuesa5_qualified_rate#1 / 1877.55Source ↗official
UAVBench safety-critical decision recognitionethical_safety_critical_accuracy#16 / 270.68Source ↗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-evalRankValueDistributionSource
SpeechMap model completioncomplete_pct#143 / 18140.3Source ↗official