返回文章列表
Models12 min
OpenAI o1/o3 vs Claude Opus Extended Thinking: Reasoning Bench
Misaki HayashiAI Research Analyst2026-04-2112 min
o1o3Extended ThinkingReasoningBenchmarks
本文以日语发表。中文摘要如下：
OpenAI o1/o3 vs Claude Opus Extended Thinking: Reasoning Bench—OpenAI o1/o3 vs Claude Opus 4.5/4.7 Extended Thinking measured on AIME 2025, GPQA Diamond, ARC-AGI 2 and SWE-bench Verified — per-task strengths broken down.