Doubao-1.5-Thinking-Pro: benchmark results

Provider: ByteDance. Released 2025-04-17. Access: API.

Aggregate Rating 1719 ± 21, rank #260 of 10516 rated models, from 43 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LLMEval-Fair - Economics9.67Discipline Score (0-10)100
LLMEval-Fair - Education9.43Discipline Score (0-10)100
LLMEval-Fair - History9.7Discipline Score (0-10)100
LLMEval-Fair - Law9.77Discipline Score (0-10)100
LLMEval-Fair - Literature8.93Discipline Score (0-10)100
LLMEval-Fair - Medicine8.97Discipline Score (0-10)100
LLMEval-Fair - Military8.97Discipline Score (0-10)100
LLMEval-Fair - Overall93.67Absolute Score (0-100)100
LLMEval-Fair - Science9.23Discipline Score (0-10)100
MedErrBench (Chinese) - Error Detection75Detection accuracy (%): whether the note contains an error (100
MedErrBench (Chinese) - Error Localization73.5Localization accuracy (%): naming the sentence that contains100
MedErrBench (English) - Error Correction76.6Correction ROUGE-1 (%): unigram overlap of the corrected sen100

Interactive version: theaggregate.ai/model?slug=doubao-1-5-thinking-pro · How It Works · Data refreshed daily, snapshot 2026-11-21.