新基准测试考察编程智能体长时程工程任务能力
想看智能体除了刷 LeetCode 还能不能干真活,这个基准直接考核需求收集、协作、写码、自检全流程。目前 Opus 5.5 排第一,GPT-6 Astra 紧追。选模型做工程自动化的人可以看榜单再决定。
一个新的编程智能体基准测试发布,评估智能体能否作为可靠的数字工程师完成长时程任务。通过标准覆盖五个环节:收集需求、与同事协作、编写代码、验证工作、汇报结果。当前榜单上 Opus 5.5 排名第一,GPT-6 Astra 紧随其后。
该测试与常见的单轮代码生成评测不同,重点在于多步骤、长时间跨度下的工程可靠性,要求智能体在整个工作流中自主推进并保证质量。这种设计更接近真实软件团队中工程师的日常工作方式。
对开发者而言,榜单提供了在复杂工程场景下选择模型的参考;对企业来说,能否胜任长时程任务直接影响智能体在真实研发流程中的落地价值。
New benchmark for coding agents.
This evaluates whether agents can be dependable digital engineers on long-horizon tasks.
To pass, agents must:
• Gather requirements • Coordinate with colleagues • Write the code • Verify their work • Report their results
The leaderboard currently has Opus 5.5 at the top, with GPT-6 Astra a close second.
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include coding agents, Opus 5.5, GPT-6, benchmarks.
Source attribution: curated from Ate-a-Pi (@svpino). Cross-source corroboration: 7 independent sources have covered this development (including arXiv cs.AI, Greg Brockman, IT之家), confirming industry-wide signal strength.