--- title: Jev for Python engineers | daily.dev description: Vercel introduces Jev, a new classifier-style AI model accessed through its Python AI SDK's experimental evaluate() API. Jev takes structured data and... canonical: https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Jev for Python engineers | daily.dev og:description: Vercel introduces Jev, a new classifier-style AI model accessed through its Python AI SDK's experimental evaluate() API. Jev takes structured data and... og:url: https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r og:image: https://api.daily.dev/og/posts/bBitBSB0R.png og:image:alt: Jev for Python engineers og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Jev for Python engineers **[Vercel](https://daily.dev/sources/vercel)** · 8 min read · 0 upvotes · 0 comments ## Summary Vercel introduces Jev, a new classifier-style AI model accessed through its Python AI SDK's experimental evaluate() API. Jev takes structured data and multiple-choice questions and returns answers with confidence scores, functioning like a lightweight LLM-powered classifier rather than a text generator. The author demonstrates Jev through two experiments: distinguishing Python code from English text, and attempting to generate Python code by building an AST through a series of guided choices. Both experiments reveal interesting capabilities but also notable limitations, with code and setup instructions provided for developers to try it themselves. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Jev and how does it work as an AI model? Jev is a universal classifier model that takes input data plus a set of multiple-choice questions and returns answers with confidence scores instead of generating free text. It behaves like a traditional machine learning classifier but is built on an LLM, so it does not need training for a specific domain. It is fast, cheap, and returns structured JSON output. _Developers evaluating new classifier-style AI models can track releases like this one on daily.dev._ ### How do I use the experimental evaluate() function in Vercel's AI SDK for Python to call Jev? Install the SDK with uv add ai, set an AI_GATEWAY_API_KEY from an AI Gateway key, then call ai.ops.experimental.evaluate() passing a model obtained via ai.get_model("typesafe-ai/jev"), a state string or JSON, and a mapping of questions such as ChoiceQuestion, ScoreQuestion, or NoulQuestion. The function returns a result with a value dict holding the chosen answers. _Developers wiring up new SDK APIs can follow setup walkthroughs like this one via daily.dev._ ### Can an LLM-based classifier like Jev reliably generate Python code? Not well on its own. Attempting to have Jev write Python letter-by-letter or word-by-word produced almost nothing usable, so the approach shifted to having Jev build a Python abstract syntax tree through a series of guided choices, with a separate LLM (GPT-5.6) expanding the user prompt into detailed instructions first. Even then, the generated code was syntactically valid but mostly incorrect. _Developers weighing classifier models against LLMs for code generation can compare experiences on daily.dev._ --- Tags: [#python](https://daily.dev/tags/python), [#llm](https://daily.dev/tags/llm), [#classification](https://daily.dev/tags/classification), [#jev](https://daily.dev/tags/jev) [View this post on daily.dev](https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Jev for Python engineers","url":"https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r"},"datePublished":"2026-10-02T17:10:56.706Z","dateModified":"2026-10-02T21:49:27.104Z","description":"Vercel introduces Jev, a new classifier-style AI model accessed through its Python AI SDK's experimental evaluate() API. Jev takes structured data and...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7d1f3d5496f95ec2990de075639cc070?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/7d1f3d5496f95ec2990de075639cc070?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Vercel","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Vercel","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/04cca7235cf84db092f24b7ba427b3e1","url":"https://daily.dev/sources/vercel"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"python,llm,classification,jev","timeRequired":"PT8M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Vercel","item":"https://daily.dev/sources/vercel"},{"@type":"ListItem","position":3,"name":"Jev for Python engineers"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/jev-for-python-engineers-bbitbsb0r#faq","mainEntity":[{"@type":"Question","name":"What is Jev and how does it work as an AI model?","acceptedAnswer":{"@type":"Answer","text":"Jev is a universal classifier model that takes input data plus a set of multiple-choice questions and returns answers with confidence scores instead of generating free text. It behaves like a traditional machine learning classifier but is built on an LLM, so it does not need training for a specific domain. It is fast, cheap, and returns structured JSON output. Developers evaluating new classifier-style AI models can track releases like this one on daily.dev."}},{"@type":"Question","name":"How do I use the experimental evaluate() function in Vercel's AI SDK for Python to call Jev?","acceptedAnswer":{"@type":"Answer","text":"Install the SDK with uv add ai, set an AI_GATEWAY_API_KEY from an AI Gateway key, then call ai.ops.experimental.evaluate() passing a model obtained via ai.get_model(\"typesafe-ai/jev\"), a state string or JSON, and a mapping of questions such as ChoiceQuestion, ScoreQuestion, or NoulQuestion. The function returns a result with a value dict holding the chosen answers. Developers wiring up new SDK APIs can follow setup walkthroughs like this one via daily.dev."}},{"@type":"Question","name":"Can an LLM-based classifier like Jev reliably generate Python code?","acceptedAnswer":{"@type":"Answer","text":"Not well on its own. Attempting to have Jev write Python letter-by-letter or word-by-word produced almost nothing usable, so the approach shifted to having Jev build a Python abstract syntax tree through a series of guided choices, with a separate LLM (GPT-5.6) expanding the user prompt into detailed instructions first. Even then, the generated code was syntactically valid but mostly incorrect. Developers weighing classifier models against LLMs for code generation can compare experiences on daily.dev."}}]} ```