返回全部动态
Optima 平台允许用户用自有数据定制 AI 基准测试
原标题：Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data
AI 摘要
Artificial Analysis 发布了新平台 Optima，允许用户基于自己的数据或场景描述构建定制化 AI 基准测试，并在质量、单任务成本和耗时上比较不同模型。该平台旨在解决通用基准测试无法反映实际应用场景的问题，但定制基准的有效性仍取决于设计质量。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data Key Points - Artificial Analysis has released Optima, a platform that lets users build their own AI benchmarks tailored to specific use cases. - Tests can draw on users' own data sources or descriptions of the desired scenario, complete with sample inputs and outputs. Optima then compares AI models not just on quality but also on cost per task and time per task. - The platform addresses the limited
发布时间：2026-08-16 13:50
抓取时间：2026-08-16 14:27
来源机构：THE DECODER