Perplexity将Deep Research移至Computer，调度20+前沿模型处理研究子任务，生成报告、演示文稿和仪表板
英文摘要
Perplexity integrated its Deep Research mode into Computer, the company’s multi-model orchestration system. The upgraded feature automatically breaks complex questions into subtasks and routes them across more than 20 frontier models. It uses Search as Code to generate code that runs thousands of parallel retrieval steps, dramatically improving agentic browsing: the BrowseComp benchmark score rose from 40.7% to 83.8%, and Humanity’s Last Exam rose from 36.4% to 50.5%. The system reads user-uploaded files alongside live web sources, cites every claim inline, and delivers finished reports, slide decks, and interactive dashboards. Developers can access the same search stack via the pay-as-you-go Perplexity Agent API with a deep-research preset.
中文摘要
Perplexity将Deep Research模式整合到其多模型调度系统Computer中。升级后，它能自动将复杂问题拆分为子任务，并在20多个前沿模型之间分配执行。系统采用“代码化搜索”生成代码，并行运行数千个检索步骤，大幅提升代理浏览能力：BrowseComp基准得分从40.7%跃升至83.8%，Humanity’s Last Exam从36.4%升至50.5%。它能同时读取用户上传文件和实时网页，对每个声明都进行内联引用，并直接输出成品报告、幻灯片和交互式仪表板。开发者可通过按量付费的Perplexity Agent API使用同一技术栈，并配备deep-research预设。
关键要点
Deep Research now lives inside Computer, which breaks questions into subtasks and routes them across 20+ frontier models.
Deep Research已嵌入Computer，系统可将问题拆分为子任务，并分配给20多个前沿模型。
Search as Code writes retrieval scripts that execute thousands of parallel steps per question, rather than using a fixed pipeline.
代码化搜索会编写检索脚本，每个问题可并行执行数千个步骤，而非固定流水线。
Agentic browsing accuracy improved dramatically: BrowseComp from 40.7% to 83.8%, Humanity’s Last Exam from 36.4% to 50.5%, DeepSearchQA from 81.9% to 85.0%.
代理浏览准确率大幅提升：BrowseComp从40.7%升至83.8%，Humanity’s Last Exam从36.4%升至50.5%，DeepSearchQA从81.9%升至85.0%。
The system reads internal files and live web data, cites every claim, and outputs reports, decks, and dashboards directly.
系统可读取内部文件与实时网页，每条声明均注明出处，并直接输出报告、演示文稿和仪表板。
Developers can use the same agentic search stack via the Perplexity Agent API with a pre-configured deep-research preset.
开发者可通过Perplexity Agent API使用同一代理搜索技术栈，并使用预配置的deep-research预设。