Jev introduces a new shape of LLM—System One, aka Decision Models
https://simonwillison.net/2026/Sep/21/jev/📌 Jev：不回你文字，只丟一個浮點數的新型 LLM
TL;DR：TypeSafe AI 推出 Jev，把 LLM 輸出換成機率分數，定價只算輸入 token。
如果一個模型不回答你任何一個字，只回一個浮點數，你還敢說它是 LLM 嗎？TypeSafe AI 上週發布的 Jev，就是這麼一個反直覺的東西，而它引發的討論熱度，讓 Simon Willison 特地寫了一篇文章來拆解。
🤔 一種新形狀的模型：System One，或者說決策模型
TypeSafe 把 Jev 稱為「System One models」，Simon Willison 則認同 Maggie Appleton 的說法，覺得「decision models（決策模型）」是更好的命名。Jev 依然接受文字輸入，但輸出不再是文字，而是對應到分類、是非題、評分與信心分數的浮點數。TypeSafe 自己的形容是：把它想成一個「前沿智慧等級的 function call」——丟進非結構化的 state，吐出有型別的機率決策。
🧩 怎麼用：一個 state 物件配上幾個問題
使用方式是組出一個「state」物件，內容可以是一個字串、字串陣列，或一組 name-value pairs，用來描述一篇文章、一位客戶或任何一筆紀錄。接著把這個 state 連同一個或多個問題送進 API，每個問題都會得到各自的回覆。Jev 支援三種類型的問題，而且問題是平行評估的，所以丟很多題目跟丟一題花的時間差不多。
定價上，Jev 只針對輸入收費，輸出完全免費，第一代模型的輸入價格是每百萬 token 0.042 美元，比 OpenAI GPT-5 Nano 的 0.05 美元還便宜。
📊 強項與弱項：Jev 1.13 的 jaggedness 文件說了什麼
官方的 Jev 1.13 jaggedness 文件坦承，Jev 目前不擅長處理數字、日期，以及「adversarial content（對抗性內容）」。它適合的是能表達成分類任務的工作，例如垃圾郵件偵測、建議標籤、優先順序排序。Simon 自己也拿它做搜尋 reranking 實驗：先用 BM25 這類便宜演算法抓出 100 個候選結果，再讓 Jev 針對原始查詢逐一評分相關性。
💡 比 LLM 更黑箱：只給分數，不給理由
Simon 對 Jev 感到不太自在的一點是，它比一般 LLM 更往黑箱機器學習系統倒退。一般 LLM 至少還能要求它解釋自己的判斷（即使解釋不保證準確），但 Jev 連這一步都省了：丟進再多文字，拿回來的就只有一個浮點數。如果 Jev 把某段內容判成 spam，到底是哪個訊號觸發的，完全無從得知。這也讓偏見問題變得格外重要，Simon 特別提到千萬別拿 Jev 去替求職者排名，那個浮點數背後可能藏著模型未被察覺的偏見，事後要拆解出來會很棘手。他自己做過一個實驗，讓 Jev 對舊金山灣區每個城市回答「是不是 Good city？」的是非題，結果 Cupertino 排最高、East Palo Alto 排最低。
⚠️ 便宜到能跑上千次實驗，但也代表評測不能少
正因為這些疑慮，evals 與結構化實驗在 Jev 專案裡比一般 LLM 專案更重要。好在 Jev 夠便宜，跑個幾百甚至幾千次實驗性 prompt 也只要幾分錢，這讓大規模驗證變得可行，只是能不能真的去做，取決於團隊有沒有把這一步排進流程。
🎯 實務啟示
Jev 釋出還不到一週，社群已經出現大量創意應用，也出現了用開放權重模型仿造 Jev 的專案：Kev 用 Qwen 3.5 打造出 0.8B、4B、9B 三種規模的版本，還有專門比較「Jev-class 決策模型」的 JevBench 基準測試出現在 Hacker News 討論串裡。如果你的場景是分類、排序或 reranking，且輸出本來就不需要自然語言解釋，Jev 這類決策模型可能比一般 LLM 更划算；但只要牽涉到會影響人的決策（招募、信用評分等），黑箱程度加上潛在偏見風險，都值得在導入前先想清楚。
🔗 來源
- 標題：Jev introduces a new shape of LLM—System One, aka Decision Models
- 作者／機構：Simon Willison
- 連結：https://simonwillison.net/2026/Sep/21/jev/
#Jev #DecisionModels #LLM #MachineLearning #AIBias #Classification #SearchReranking #TypeSafeAI #OpenWeights #AIEvals
原始資料 Simon Willison · 收集於 2026-09-22
摘要原文
Last week TypeSafe AI unveiled Jev , their first example of a new category of model that they are calling “System One models” (I’m with Maggie Appleton, I think “decision models” is a better name for these). Jev is an interesting variant on the usual LLM format: it still accepts text inputs, but instead of text output it returns floating point numbers corresponding to categories, yes/no questions, ratings, and associated confidence scores. TypeSafe describe Jev like this: Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out. It’s also very fast, and really cheap . Regular LLMs are priced in terms of input and output tokens, with output generally charged at significantly higher rates. Jev charges only for input—output is free—and the input price of their first model is $0.042 per million tokens—cheaper even than OpenAI’s GPT-5 Nano ($0.05/million). Jev lets you ask questions about text or semi-structured data. You compose a “state” object containing a string, array of strings, or set of name-value pairs—this might describe an article, or a customer, or any other kind of record. You then send that to their API with one or more questions, and get a reply back for each. You can ask three kinds of questions: The Jev API can accept a single document (“state”) and as many questions as you can cram into the context window. Questions are evaluated in parallel, so sending many questions should take a similar time to sending just one. The Jev 1.13 jaggedness documentation offers useful guidance as to Jev’s strengths and weaknesses. It’s currently not great with numbers, dates, or “adversarial content”. I think the decision model framing is useful for understanding where to use Jev. It’s great for anything that can be expressed as a classification task—think spam detection, suggesting labels, prioritization and ranking. I’ve also been experimenting with it for search reranking, where you fetch 100 likely matches using an inexpensive algorithm like BM25, then have Jev score those 100 candidates for relevance against the original query. Something I’ve found a little uncomfortable about Jev is how it very much represents a regression even further towards black box machine learning systems. LLMs are black boxes already—you can ask them to justify their decisions, but you can’t guarantee that what they say is useful or accurate. Jev doesn’t even give you that: put in all the text you want, the only thing you’re going to get back is a floating point number. If Jev marks something as spam, which content signals tipped it off? This also means that concerns about bias should be front and center. I really hope nobody uses Jev to rank job applicants—that floating point number could conceal all manner of unseen bias baked into the models, and experimentally picking that bias apart is going to be a tricky business. (I tried one experiment where I had Jev score every city in the San Francisco Bay Area on a yes/no answer to whether they were a “Good city?”—it rated Cupertino top and East Palo Alto bottom. Huh.) In practice, this all means that evals and structured experiments are even more important than they are for regular LLM projects. Thankfully, Jev is so cheap that running hundreds or even thousands of experimental prompts through it costs just a few cents. It’s been really fun watching the wider community come up with potential use-cases for Jev over the past few days. Here are some creative ones that caught my eye: There’s also been a flurry of projects attempting to create a model like Jev using on top of open weight models. Kev is one interesting example, using Qwen 3.5 to produce 0.8B, 4B, and 9B models. Here’s the accompanying Hacker News thread , where someone linked to a JevBench benchmark that has already cropped up to compare “Jev-class decision models”. Given Jev was released just under a week ago, the amount of activity around it is extremely impressive. This is Jev introduces a new shape of LLM—System One, aka Decision Models by Simon Willison, posted on 21st September 2026 . Previous: Generating running routes with GPT-6 Astra and ChatGPT Work Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.
由 claude-code/sonnet 自動生成