Benchmark Almanac

ERNIE-4.5-300B-A47B

ERNIE-4.5-300B-A47B ranks first on 2 of the 29 benchmarks with five or more models scored.

Baidu Released June 30th, 2025 Family ERNIE 4.5 The lab's model card Compare with others
30
Benchmarks with a score
2
Ranked first, out of five or more models
6
In the top three
Baidu Jun 2025

ERNIE-4.5-300B-A47B

Open weights Flagship

126.9 #302 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • FullStackBench 67.1% #3 /8
  • SysBench 69.8% #5 /6
  • MMLU-Pro 78.4% #90 /167

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
BIG-Bench Hard 94.3% #1 of 73 Lab model card
CMMLU 90.2% #3 of 17 Lab model card
DROP 91.1% #8 of 41 Lab model card
SimpleQA 45.9% #15 of 60 Lab model card
Chinese SimpleQA 77.1% #4 of 13 Lab model card
Multi-IF 76.6% #9 of 28 Lab model card
MuSR 69.9% #2 of 6 Lab model card
MMLU 86.5% #45 of 131 Lab model card
C-Eval 90.6% #16 of 46 Lab model card
IFEval 88% #53 of 135 Lab model card
LLM Confabulation Leaderboard 16% #28 of 58 Official leaderboard
MMLU-Pro 78.4% #90 of 167 Lab model card
SysBench 69.8% #5 of 6 Lab model card
BullshitBench 12.7% #93 of 107 Official leaderboard
BullshitBench V2 4.00% #126 of 129 Official leaderboard

Math and science

BenchmarkScoreRankSource
GSM8K 96.6% #2 of 92 Lab model card
CMATH 96.7% #1 of 6 Lab model card
MATH-500 96.4% #28 of 70 Lab model card
AIME 2024 54.8% #73 of 88 Lab model card
AIME 2025 35.1% #136 of 143 Lab model card
ZebraLogic 58.1% #22 of 23 Lab model card
OlympicArena 75.3% #1 of 1 Lab model card

Coding

BenchmarkScoreRankSource
HumanEval+ 92.1% #6 of 27 Lab model card
MBPP+ 81% #4 of 15 Lab model card
FullStackBench 67.1% #3 of 8 Lab model card
LiveCodeBench 38.8% #35 of 54 Lab model card
LiveCodeBench v6 38.8% #87 of 90 Lab model card

Agents and tools

BenchmarkScoreRankSource
SnakeBench 23 #104 of 188 Official leaderboard
LisanBench 176.3 #94 of 133 Official leaderboard

Other

BenchmarkScoreRankSource
LiveBench 68.9% #13 of 27 Lab model card

Its lab's numbers put it at 126.9, number 302 of 404

The Self-Reported Index fits one score to the 20 capability results Baidu reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 124.1 to 129.7. How we compute it.

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
MMLU-Pro 78.4% MMLU-Pro 78.4% Leaderboard Same

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Its system card

Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.