Replit· @Replit · X·· 8 小时前AI 评分
Replit 发布 ViBench 2.0 长程软件构建基准
Replit 发布 ViBench 2.0，用于评估 AI 智能体从空项目开始逐个功能构建完整产品的能力，测试过程中不会预先看到后续任务。GPT-6.1 Sol、GPT-6 Astra 和 Opus 5.5 等模型仅在 49% 至 65% 的构建中通过全部检查。该基准已纳入 FireworksAI 的 Specialized Intelligence Index，并收录 Fireworks 的官方运行结果。
Today we're releasing ViBench 2.0, a benchmark for long-horizon software building. Agents start from an empty project and build a whole product, one feature at a time, without seeing what comes next.
Even the best models (GPT-6.1 Sol, GPT-6 Astra, Opus 5.5) pass every check in only 49-65% of builds.
ViBench 2.0 is now part of @FireworksAI_HQ’s Specialized Intelligence Index, with official results from Fireworks’ runs.
Read more: replit.com/blog/vibench-2-lo…
来源：Replit · x.com