**Show HN: Portal2Bench – Can AI agents solve Portal 2?**

> Can AI agents solve Portal 2? GPT-6.1 Sol solves 20 of 26 test chambers and GPT-6 Astra solves 20, each within $30 and 300 plans. Portal2Bench Portal2Bench measures whether AI agents can solve Portal 2&\#x27;s test chambers. Each chamber is a fresh, independent run, scored solved only if the agent finishes within $30 and 300 plans. Read research Soon Get in touch Leaderboard Model \* Agent harness Chambers solved † GPT...

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：论文
- **发布时间**：2026-11-14 08:40（北京时间）
- **原文**：[打开原文](https://kingstonyip.com/portal2bench)