字节跳动：2025年思考模型Seed-Thinking-v1.5技术报告（英文版）

We introduce Seed1.5-Thinking, capable of reasoning through thinking before responding, resultingin improved performance on a wide range of benchmarks. Seed1.5-Thinking achieves 86.7 onAIME 2024. 55.0 on Codeforces and 77.3 on (;P()A, demonstrating excellent reasoning abilities inSTEM and coding. Beyond reasoning tasks, the method demonstrates notable generalization acrossdiverse domains. For instance, it surpasses DeepSeek Rl by 8% in win rate on non-reasoning tasksindicating its broader app