Exa Launches Agent Ultra: A Subagent Swarm Deep Research API Built for Exhaustive List Building
https://www.marktechpost.com/2026/09/26/exa-launches-agent-ultra-a-subagent-swarm-deep-research-api-built-for-exhaustive-list-building/📌 Exa推出Agent Ultra，把研究做到「窮盡」為止
TL;DR：Exa發布Deep Research API最高檔位Agent Ultra，宣稱在4項benchmark上超越Opus 5.5、GPT-6 Astra與Perplexity Agent。
如果一個研究任務要花上30分鐘、甚至3小時才跑完,還得查證上千筆來源,一般agent恐怕早就見好就收了。Exa偏偏把「不查到窮盡不罷休」做成了一個產品檔位。
🤔 為窮盡式清單建構而生
Exa Agent Ultra是Exa Agent API目前效力最高的模式,鎖定的是那些必須「跑到窮盡」的研究任務:大規模清單建構、實體資料補全,以及需要動用上千筆來源才能回答的問題。
🧩 把任務拆給一群子代理去跑
Exa Agent的做法是把一個任務拆解成多個子任務,交給不同子代理同時研究多個領域;在需要的步驟上調度前沿模型,其餘步驟則交給速度較快的模型處理。Ultra是其中花費運算資源最多的檔位,官方文件指出它會比其他效力等級跑得更久,以換取最完整的結果——一般複雜任務約需30分鐘,難度極高的任務則可能耗時到3小時。
目前Agent Ultra已經以託管API形式上線,只要在請求中設定effort: “ultra”即可呼叫,並非開放權重模型,也無法自行部署(self-host)。它沿用標準的Agent run端點,支援outputSchema、input.data與串流輸出;如果手上已經有一份清單,也可以把既有資料傳入,讓Ultra只補上新結果、排除重複項目。使用者可以先在Exa API Playground中試跑。
📊 4項benchmark的比較數字
根據Exa發布的資料,Agent Ultra在WANDR、DeepSearchQA、WideSearch與Company Find-All這4項研究benchmark上,分別對比了同樣以最高效力設定運行的Opus 5.5、GPT-6 Astra與Perplexity Agent。WANDR是Perplexity提出的benchmark,含500個廣度與深度兼具的資料蒐集任務,附有開放的評測框架(harness);DeepSearchQA則是Google DeepMind的900道多步驟搜尋題;WideSearch測試的是廣泛資訊蒐集能力。Exa在WANDR與DeepSearchQA上各評測了最多200個任務,WideSearch與Company Find-All則各評測100個,不同廠商實際被評分的任務數量並不相同。在WANDR上,Ultra與Opus 5.5的絕對分差是9.1分。凡是廠商已在該評測框架上公開過成績,Exa便直接引用;其餘情況則由Exa自行跑測。
⚠️ 以上所有數字均來自Exa自家的發布文章與benchmark說明,屬廠商自報成績,尚未經過獨立第三方覆現驗證,且各家被評測的任務數量本身就不完全一致,比較時需留意這層落差。
🎯 實務啟示
如果團隊本身就有大規模清單建構或實體資料補全的需求,Agent Ultra這種「用運算換完整度」的檔位值得評估,但因為是純託管API、無法自行部署,也要一併考量資料是否適合交給第三方服務處理。面對廠商自報的benchmark數字,建議先用小規模真實任務驗證再決定是否全面導入,尤其是動輒30分鐘到3小時的執行時間,不適合放進對延遲敏感的流程裡。
🔗 來源
- 標題：Exa Launches Agent Ultra: A Subagent Swarm Deep Research API Built for Exhaustive List Building
- 作者／機構：Michal Sutter／MarkTechPost
- 連結：https://www.marktechpost.com/2026/09/26/exa-launches-agent-ultra-a-subagent-swarm-deep-research-api-built-for-exhaustive-list-building/
#Exa #AIAgents #DeepResearch #LLM #AgentSwarm #Benchmark #AISearch #DataEnrichment #API #MachineLearning
原始資料 MarkTechPost · 收集於 2026-09-27
摘要原文
Exa has released Agent Ultra , the highest effort level of its Exa Agent API. It is built for research that must run to exhaustion: large list building, entity enrichment, and questions that need thousands of sources. Exa team reports that Ultra beats Opus 5.5, GPT-6 Astra, and Perplexity Agent, each at maximum effort, on 4 research benchmarks. Is it deployable? Yes, as a hosted API. Agent Ultra is live today on the Exa API by setting effort: "ultra" . It is not open weights and cannot be self-hosted. Exa Agent splits a task into subtasks and assigns subagents to research several domains at once. It routes frontier models to steps that need them and faster models where those are enough. Ultra is the mode that spends the most compute. According to the Agent Ultra docs , it runs longer than any other effort to return the most complete results. Ultra runs typically finish complex tasks in about 30 minutes. Very hard tasks can take up to 3 hours. All figures below are from Exa’s launch post . Competitors ran at their maximum effort setting. Exa pairs each result with a cost claim: These gains are relative, not percentage points. On WANDR, the absolute gap to Opus 5.5 is 9.1 points. WANDR is Perplexity’s benchmark of 500 wide and deep data-collection tasks, with an open harness . Exa’s grader shares the upstream evaluation logic. It swaps in Exa as the contents tool, changes transport logic, and uses gpt-6-luna as the judge. Where a vendor had published a result on this harness, Exa reports that figure. Otherwise, Exa ran the benchmark itself. DeepSearchQA is Google DeepMind’s 900-prompt multi-step search benchmark. WideSearch tests broad information gathering. Exa evaluated up to 200 tasks each for WANDR and DeepSearchQA, and 100 each for WideSearch and Company Find-All. Graded task counts vary by provider. All results are vendor-reported and not yet independently reproduced. Exa lists 3 target user groups: Ultra can also expand an existing list. Pass the rows you already have, and they are excluded from new results. Ultra uses the standard Agent run endpoint. The request supports outputSchema , input.data , and streaming. You can test it in the Exa API Playground . Check out the Technical Details . All credit goes to the researcher of this project. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter . Wait! are you on telegram? now you can join us on telegram as well. Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us The post Exa Launches Agent Ultra: A Subagent Swarm Deep Research API Built for Exhaustive List Building appeared first on MarkTechPost .
由 claude-code/sonnet 自動生成