Analyzing Jev, a new AI model

Analyzing Jev, a new AI model解析新型 AI 模型 Jev

September 22, 20262026 年 9 月 22 日

A new AI model called Jev launched last week and went viral.一款名为 Jev 的新型 AI 模型于上周发布，并迅速走红。

It's not an LLM. It doesn't generate text or pictures, but could be very useful for a number of use cases. The creators call it “the first System One model”, tout its performance (193x faster, 444x cheaper) and promise it can't hallucinate.它不是大语言模型（LLM）。它不生成文本或图像，但在多种应用场景中可能非常有用。其开发者称之为“首个系统 1 模型”，并吹捧其性能（速度提升 193 倍，成本降低 444 倍），同时承诺它不会产生幻觉。

So, what is it?那么，它到底是什么？

How it (probably) works它是（可能）如何工作的

Behind the impressive marketing is a very interesting model. Jev is a close relative of LLMs like GPT, Claude, or DeepSeek, but is actually more similar to the BERT series of models, which are used for classification, not generation. (If you've ever wondered why someone calls out a difference between “AI” and “GenAI”, that's the distinction they mean).在令人印象深刻的营销背后，是一个非常有趣的模型。Jev 是 GPT、Claude 或 DeepSeek 等大语言模型的近亲，但实际上它更类似于用于分类而非生成的 BERT 系列模型。（如果你曾好奇为什么有人区分“AI”和“生成式 AI”，这就是他们所指的区别。）

TypeSafe (the creators of Jev) haven't actually published any papers on the system, but here's a reasonable guess and simplification, just to give you a sense of how it works.TypeSafe（Jev 的开发者）实际上并未发布过任何关于该系统的论文，但为了让你了解它的工作原理，这里提供一个合理的猜测和简化说明。

Imagine going back to 2022. GPT-3, the best LLM at the time, can autocomplete text (you give it some text as an input and it continues it). Then InstructGPT shows how to make the LLM follow instructions from the user. Following the ChatGPT launch, the advances pile up, with reasoning, reinforcement learning, agentic use cases, and all the modern tech tricks LLMs now employ.想象一下回到 2022 年。当时最优秀的大语言模型 GPT-3 可以自动补全文本（你输入一段文本，它接着写下去）。随后，InstructGPT 展示了如何让大语言模型遵循用户的指令。随着 ChatGPT 的发布，技术进步不断积累，引入了推理、强化学习、智能体用例以及大语言模型现在采用的所有现代技术技巧。

BERTs, meanwhile, are almost stuck in time. There's ModernBERT, which modernizes it a lot, but you still need to fine-tune your model for the specific task. This is annoying, cumbersome, and more expensive than just asking an LLM. But these classification models are still much faster and cheaper than LLMs for suitable tasks, once you go through the trouble of training them.与此同时，BERT 系列模型几乎停滞不前。虽然有 ModernBERT 对其进行了大量现代化改造，但你仍然需要针对特定任务对模型进行微调。这既烦琐又麻烦，而且比直接询问大语言模型更昂贵。但一旦你完成了训练，这些分类模型在处理合适任务时，速度和成本仍然远优于大语言模型。

Now imagine someone going, “OK, how can we modernize BERTs with all the cool tech LLMs have been getting? Like using instructions in the input instead of having to re-train, making it really large, and using reinforcement learning?” That's basically what Jev is (not exactly right, but it's a good analogy).现在想象一下有人提出：“好吧，我们如何利用大语言模型获得的所有炫酷技术来升级 BERT？比如在输入中使用指令而不是重新训练，将其规模做大，并使用强化学习？”这基本上就是 Jev 的本质（虽然不完全准确，但这是一个很好的类比）。

Since it only ingests the input and provides a few probabilities as output, it's much faster and cheaper than LLMs. Jev's model size is not publicly known, but even if it's 100x larger than ModernBERT, it'd still be a paltry 40B params next to the humongous 2T-5T LLMs.由于它只接收输入并输出几个概率值，因此它比大语言模型快得多且便宜得多。Jev 的模型规模尚未公开，但即使它比 ModernBERT 大 100 倍，与 2T-5T 参数的庞大语言模型相比，也仅仅是 40B 参数的微小规模。

The authors also mention performance improvements due to not needing to generate text tokens and other workflow simplification, as well as the fact that they can evaluate the questions in parallel.开发者还提到，由于无需生成文本 Token 以及简化了工作流程，性能得到了提升，同时他们还可以并行评估问题。

The “System One” moniker is a clever marketing trick that means “it's not a reasoning model”. They borrow the term from Thinking, Fast and Slow, where System 1 means intuitive, fast, automatic thinking and System 2 means deliberate, slower, effortful reasoning.“系统 1”这个称号是一个聪明的营销技巧，意指“它不是一个推理模型”。他们借用了《思考，快与慢》一书中的术语，其中“系统 1”代表直觉、快速、自动化的思维，“系统 2”代表审慎、缓慢、费力的推理。

What about the hallucination claim? In the context of Jev, no hallucination means it can't generate output outside of the specified structure (i.e. it's “type-safe”), but that doesn't mean it can't generate incorrect output.那么关于不会产生幻觉的说法呢？在 Jev 的语境下，没有幻觉是指它无法生成超出指定结构之外的输出（即它是“类型安全”的），但这并不意味着它不会生成错误的输出。

Using Jev使用 Jev

This brings me to the usage part: you'd use Jev (and other classification models like BERTs) to answer very specific questions, like “is this email spam?”, and it can answer with a Yes/No, a score (from a continuous domain, like 0.0-10.0), or a choice from a few discrete possibilities (classification).这引出了使用部分：你可以使用 Jev（以及 BERT 等其他分类模型）来回答非常具体的问题，例如“这封邮件是垃圾邮件吗？”，它可以回答“是/否”、一个分数（来自连续域，如 0.0-10.0）或从几个离散选项中进行选择（分类）。

Interestingly, the questions are independent and evaluated in parallel, which helps with performance. It also means you can just batch all the questions/classifications for a given piece of data at once, instead of going step by step.有趣的是，这些问题是独立的并可并行评估，这有助于提升性能。这也意味着你可以一次性批量处理给定数据的所有问题/分类，而无需一步步进行。

Since Jev doesn't need to be fine-tuned, it's easy to use it via an API (and in fact, it's the only way to use it – don't expect the weights). Here's one official example, in Python:由于 Jev 不需要微调，因此通过 API 使用它非常简单（事实上，这也是使用它的唯一方式——别指望能拿到模型权重）。以下是一个官方的 Python 示例：

from typesafe_sdk import Choice, Noul, Score, TypeSafeClient

client = TypeSafeClient()

ticket = """
Hi, I've been trying to connect my Stripe account for 3 days
and the integration keeps failing.
I'm losing sales. Please help ASAP.
"""

response = client.system_one(
    state=ticket,
    questions={
        "department": Choice(
            instructions="Which team should handle this",
            criteria={
                "billing": "Payment or subscription issues",
                "technical": "Bugs or integration problems",
                "sales": "Pricing or account questions",
            },
        ),
        "frustration": Score(
            instructions="How frustrated the customer appears",
            criteria=[
                "Calm, just stating facts",
                "Frustrated but civil",
                "Very angry, strong language",
            ],
        ),
        "is_urgent": Noul(
            instructions="The message conveys urgency or time-sensitivity",
        ),
    },
)

print(response.answers["department"].choice)  # "technical"
print(response.answers["frustration"].score)  # 1.0
print(response.answers["is_urgent"].noul)     # 1.0

Here you describe the input (i.e. the “state” of the system) and ask a number of questions. The instructions on how to classify/decide are actually part of each question, not the input state.在这里，你描述输入（即系统的“状态”）并提出若干问题。关于如何分类/判定的指令实际上是每个问题的一部分，而不是输入状态的一部分。

How about prompt injection?提示词注入（Prompt Injection）怎么样？

If this is not a quirk of the API but actually how the model was trained, with state and instructions being separate, this might mean Jev is much harder to prompt-inject.如果这不是 API 的特性，而是模型实际的训练方式（状态和指令分离），这可能意味着 Jev 更难受到提示词注入攻击。

I tried a few times to convince the above example to route the query to a different department but was unsuccessful. While I might have had more success had I tried harder, it does seem like Jev is a bit more resilient to these kinds of attacks.我尝试了几次诱导上述示例将查询路由到不同的部门，但没有成功。虽然如果我更努力尝试可能会成功，但 Jev 看起来确实对这类攻击更具韧性。

The authors are coy about this aspect, saying the adversarial examples for Jev are different from those of LLMs so that they work even better together.开发者对这一方面讳莫如深，仅表示 Jev 的对抗性示例与大语言模型的不同，因此它们可以更好地协同工作。

Jev vs BERTJev 与 BERT 的对比

A few months ago I trained a BERT model to classify some Croatian language texts, and I wondered how Jev would fare on the same task, so I gave it a go.几个月前，我训练了一个 BERT 模型来对一些克罗地亚语文本进行分类，我想知道 Jev 在同一任务上的表现如何，于是我试了一下。

I ran it zero-shot on a random sample of 3,000 texts, with a one-sentence English instruction, and the (Croatian) text, asking for binary classification.我用 3000 条随机文本样本进行了零样本（zero-shot）测试，使用了一句英文指令和（克罗地亚语）文本，要求进行二分类。

Jev got it right 97.1% of the time, compared to 97.6% for my fine-tuned model – basically the same. Looking at the cases where the two models disagreed, most turned out to be mislabeled data, and Jev was the one following the stated definition more literally.Jev 的准确率为 97.1%，而我微调的模型为 97.6%——基本持平。观察两个模型结果不一致的情况，大部分是因为数据标注错误，而 Jev 是更严格遵循既定定义的那一个。

Cost and speed were as advertised: the entire run took about a minute and cost around 20 cents. My local model ran at about the same speed (on NVidia RTX 3060).成本和速度正如宣传的那样：整个运行过程大约花费了一分钟，成本约为 20 美分。我的本地模型运行速度大致相同（在 NVidia RTX 3060 上）。

Is it useful?它有用吗？

Yes. If your use case is something that fits how Jev works, it could be much better than LLMs. Of course, if you already have a locally tuned BERT, the gains may not be much, if any – should be easy to test!是的。如果你的用例符合 Jev 的工作方式，它可能比大语言模型好得多。当然，如果你已经有了一个本地调优的 BERT，收益可能并不大（如果有的话）——测试一下应该很容易！

In my case, had I had access to Jev a few months ago, I wouldn't have needed to go to trouble of fine-tuning my own model. I'd spend a tiny bit more money (ignoring the hardware cost), which would be more than recouped in development savings.就我而言，如果几个月前我能用上 Jev，我就不需要费心去微调自己的模型了。我会多花一点点钱（忽略硬件成本），但这在开发成本的节省上会得到充分补偿。

What I'm most excited about is the combination of technical excellence and marketing excellence that raised the profile for this type of AI system, which has fallen by the wayside compared to the hype and capabilities of LLMs. That led people to use them in cases where they really aren't the best fit.我最兴奋的是技术卓越性与营销卓越性的结合，这提升了这类 AI 系统的知名度。相比于大语言模型的炒作和能力，这类系统一直被边缘化，导致人们在并不适合的场景中过度使用大语言模型。

Jev and all the copycats that are sure to appear will bring a breath of fresh air and some extra breathing room to many orgs' wallets :)Jev 以及肯定会出现的模仿者，将带来一股清新的空气，并为许多组织的钱包提供一些额外的喘息空间 :)