ERNIE 5.0 Thinking Preview

Baidu

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0 / $0

in / out

Context

—

Avg score

53.4

10 benchmarks

AA Intelligence

21.9

index

Weights

Open

Released

Nov 13, 2025

About

ERNIE 5.0 Thinking Preview is a open-weights model in the ERNIE family from Baidu. Benchmarked on 10 evals, averaging 53.4.

Benchmark scores · 10

avg 53.4 — higher is better

AIMEartificial-analysis · 2026-07

τ²-benchartificial-analysis · 2026-07

83.9

MMLU-Proartificial-analysis · 2026-07

LiveCodeBenchartificial-analysis · 2026-07

81.2

GPQA Diamondartificial-analysis · 2026-07

77.7

IFBenchartificial-analysis · 2026-07

41.4

SciCodeartificial-analysis · 2026-07

37.5

Terminal-Bench Hardartificial-analysis · 2026-07

Humanity's Last Examartificial-analysis · 2026-07

12.7

AA Long-Context Reasoningartificial-analysis · 2026-07

6.7

Details

PublisherBaidu

FamilyERNIE

Parameters—

License—

WeightsOpen

ReleasedNov 13, 2025

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0/M