Doubao-Seed-1.6-thinking-250715: benchmark results
ByteDance's Doubao-Seed-1.6-thinking (July 15, 2025 checkpoint), a 230-billion-parameter, 23-billion-active MoE reasoning model using Adaptive Chain-of-Thought to scale effort to question difficulty. Provider: ByteDance. Released 2025-07-15. Access: API.
Aggregate Rating 1717 ± 19, rank #169 of 1635 rated models, from 84 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenCompass Agent - Hard Multi-Turn (CompassBench 2507) | 58 | Score (%) | 100 |
| OpenCompass Agent - Tool Use (CompassBench 2507) | 68.7 | Score (%) | 100 |
| OpenCompass Knowledge - Humanities - English (CompassBench 2507) | 91 | Score (%) | 100 |
| OpenCompass LLM - Agent (CompassBench 2507) | 63.1 | Score (%) | 100 |
| OpenCompass LLM - Language (CompassBench 2507) | 86.8 | Score (%) | 100 |
| OpenCompass LLM - Language - Chinese (CompassBench 2507) | 86.8 | Score (%) | 100 |
| OpenCompass LLM - Math - English (CompassBench 2507) | 70.9 | Score (%) | 100 |
| OpenCompass Language - Creation - Chinese (CompassBench 2507) | 91.3 | Score (%) | 100 |
| OpenCompass Language - Instruction Following - Chinese (CompassBench 2507) | 84.8 | Score (%) | 100 |
| OpenCompass Math - Competition - English (CompassBench 2507) | 80 | Score (%) | 100 |
| OpenCompass Reasoning - Common - Chinese (CompassBench 2507) | 88.4 | Score (%) | 100 |
| SuperCLUE General (July 2025) - Agent | 90.67 | Score | 100 |
Interactive version: theaggregate.ai/model?slug=doubao-seed-1-6-thinking-250715 · How It Works · Data refreshed daily, snapshot 2026-11-05.