ERNIE-4.5-Turbo-128K: benchmark results

Provider: Baidu. Access: API.

Aggregate Rating 1621 ± 34, rank #820 of 10916 rated models, from 57 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Clarify-Then-Search (3 Questions)28.34Nugget restoration score (0-100): mean restore_score_100, we100
OpenCompass Knowledge - Social Science - English (CompassBench 2507)89Score (%)86.1
OpenCompass Knowledge - Humanities - English (CompassBench 2507)86Score (%)77.8
OpenCompass LLM - Knowledge - Chinese (CompassBench 2507)86.2Score (%)75
OpenCompass Language - Instruction Following - English (CompassBench 2507)76.4Score (%)66.7
OpenCompass Knowledge - Engineering - English (CompassBench 2507)84Score (%)61.1
OpenCompass LLM - Knowledge (CompassBench 2507)86.8Score (%)61.1
OpenCompass Language - Instruction Following - Chinese (CompassBench 2507)77.8Score (%)55.6
Clarify-Then-Search (1 Question)23.46Nugget restoration score (0-100): mean restore_score_100, we50
Clarify-Then-Search (2 Questions)26.35Nugget restoration score (0-100): mean restore_score_100, we50
OpenCompass Agent - Tool Use (CompassBench 2507)41.5Score (%)50
OpenCompass Knowledge - Science - Chinese (CompassBench 2507)86Score (%)50

Interactive version: theaggregate.ai/model?slug=ernie-4-5-turbo-128k · How It Works · Data refreshed daily, snapshot 2026-11-08.