Doubao-1.5-thinking-pro-250415: benchmark results
ByteDance's Doubao-1.5-thinking-pro (April 15, 2025 checkpoint), a 200-billion-parameter, 20-billion-active MoE reasoning model with a Deep Thinking mode, sharing its base with Seed-Thinking-v1.5. Provider: ByteDance. Released 2025-04-17. Access: API.
Aggregate Rating 1693 ± 35, rank #211 of 1640 rated models, from 60 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenCompass Language - Dialogue - Chinese (CompassBench 2504) | 95.5 | Score (%) | 100 |
| OpenCompass Language - NLP - English (CompassBench 2504) | 84.7 | Score (%) | 100 |
| OpenCompass Reasoning - Common - English (CompassBench 2504) | 90.1 | Score (%) | 100 |
| SuperCLUE General (May 2025) - Text Understanding and Creation | 81.04 | Score | 100 |
| OpenCompass Knowledge - Engineering - English (CompassBench 2504) | 92 | Score (%) | 98.1 |
| SuperCLUE General (May 2025) - Science Reasoning | 52.08 | Score | 97.4 |
| OpenCompass LLM - Average (CompassBench 2504) | 73.3 | Score (%) | 96.3 |
| OpenCompass LLM - Code - Chinese (CompassBench 2504) | 62.7 | Score (%) | 96.3 |
| OpenCompass LLM - Knowledge - Chinese (CompassBench 2504) | 89.5 | Score (%) | 96.3 |
| OpenCompass LLM - Language - English (CompassBench 2504) | 74.8 | Score (%) | 96.3 |
| OpenCompass LLM - Reasoning - English (CompassBench 2504) | 92.2 | Score (%) | 96.3 |
| OpenCompass Language - Creation - Chinese (CompassBench 2504) | 90.9 | Score (%) | 96.3 |
Interactive version: theaggregate.ai/model?slug=doubao-1-5-thinking-pro-250415 · How It Works · Data refreshed daily, snapshot 2026-10-13.