Event ·Multi-source

Agent Arena 数据：头部模型长任务耗时差距达 55 分钟

Agent Arena 公布长任务评测数据，任务覆盖数百轮对话，包括移植代码库到 Rust、审计投资策略文档、编写数学教材等。对比 5 家头部实验室的旗舰模型发现，长任务耗时差距达 55 分钟。Claude Opus 5.5 排名第一，单任务耗时最长峰值达 1 小时 51 分钟；Kimi K3.1 Pro 耗时最短，长任务仅 56 分钟。其余对比模型包括 Gemini 4 Carbon、GPT-6 Astra 和 Muse Wonder Full。

Current read

Agent Arena 拿 5 家实验室的旗舰模型跑了长任务对比，Claude Opus 5.5 能连续干近 2 小时，Kimi 最快但也最短，差距挺直观。

11 sources26° AI heatLast updated 2026/11/4 17:30:11

Evidence

11 sources

Primary sourcelmarena.ai

2026/11/4 17:30:11View source

Related source 1Artificial Analysis

2026/11/4 00:47:06View source

Related source 2rohanpaul_ai

2026/11/3 17:05:23View source

Related source 3Thomas Wolf

2026/11/4 17:04:18View source

Related source 4Decoder

2026/11/3 20:15:45View source

Related source 5marktechpost

2026/11/2 18:41:41View source

Related source 6Simon Willison’s Weblog

2026/11/2 23:49:30View source

Related source 7elvis

2026/11/3 00:32:08View source

Related source 8Gorden Sun

2026/11/3 01:43:17View source

Related source 9IT之家

2026/11/3 02:32:47View source

Related source 10shao__meng

2026/11/3 14:09:53View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story