Doubao-1.5-Pro-32k: benchmark results

Provider: ByteDance. Access: Open.

Unified ELO 1700 ± 33, rank #350 of 2928 rated models, from 23 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SuperCLUE-Faith (May 2025) - Overall95.89Accuracy (%)100
Web-Bench16.6Pass@1 (%)63.8
AGI-Eval Community - Cognition (English)80.71Accuracy (%)62.3
AGI-Eval Community - Learning (Chinese)77.14Accuracy (%)45.7
SuperCLUE-Fact - Overall81.72Score44.4
AGI-Eval Community - Cognition76.56Accuracy (%)42.5
AGI-Eval Community - Instruction Following34.89Accuracy (%)39.3
AGI-Eval Community - Cognition (Chinese)73.55Accuracy (%)37
AGI-Eval Community - Subject Reasoning (Chinese)74.73Accuracy (%)34.1
AGI-Eval Community - Subject Knowledge79.47Accuracy (%)31.4
AGI-Eval Community - Interaction (English)87.18Accuracy (%)29
AGI-Eval Community - Subject Reasoning75.79Accuracy (%)27.9

Interactive version: theaggregate.ai/model?slug=doubao-1-5-pro-32k · How It Works · Data refreshed daily, snapshot 2026-09-23.