CommunityJev Models & Reimplementations
Cloudflare Clef
Cloudflare's first open decision models on Workers AI: Clef, a 27B model that reads text, JSON, images, or video, and Clef-flash, a 9B model for the latency-sensitive hop, each returning a probability per option.
- Published by
- Community
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Capabilities
- Open source
Highlights
- Clef is a 27B multimodal model; Clef-flash is a 9B model built on a Qwen3.5-9B backbone.
- Cloudflare reports medians of 209.3 ms for Clef and 38.8 ms for Clef-flash against Jev's 524.1 ms across 43 runs.
- System One-compatible: 1 to 64 typed questions per request, with ids up to 100 characters.
- Clef adds a multimodal extension of up to four images, and the weights are open on Hugging Face.
Watch out
The latency and quality figures are Cloudflare's own, and a Hacker News tester reported the hosted model slower and weaker than Jev on one task; measure on your own questions before switching.
Questions about this entry
Short answers drawn from the linked project or page.
Cloudflare's first open decision models. Clef is a 27B multimodal model and Clef-flash a 9B model built on a Qwen3.5-9B backbone. Both answer typed questions with a probability for every allowed option and run on Workers AI.
Cloudflare reports, across 43 benchmark runs, a 209.3 ms median for Clef and 38.8 ms for Clef-flash against 524.1 ms for Jev, which makes Clef about 2.5x faster at the median and Clef-flash about 13x faster.
Yes. Clef is System One-compatible: you send a state plus 1 to 64 typed questions, question ids may use letters, digits, underscore, dot, and hyphen up to 100 characters, and answers come back under the same ids. Clef adds a multimodal extension of up to four images.
Yes. Clef and Clef-flash weights are published on Hugging Face under Cloudflare, and the models are Apache-2.0, so they can be self-hosted rather than only called through Workers AI.
Through the Workers AI binding with env.AI.run, through the REST API at /ai/run, or through AI Gateway. Cloudflare also launched a reinforcement-learning platform alongside Clef for fine-tuning decision models on your own data.
Treat them as vendor results. A Hacker News tester reported hosted Clef two to three times slower than Jev and worse on a hate-speech task, so measure on your own task before switching; the API shape is portable enough to compare on identical questions.
When you want open weights, image or video input, or a 9B option for the fastest hop, and you are willing to run your own evaluation. Jev remains the reference for text-only calibration and is the model the ecosystem benchmarks against.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Failing open or failing closed is a per-action choice, not a global default.
This post couldn't be loaded here — read it on X.https://x.com/jevdirectory/status/2781000000000000004LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
A multimodal decision model on Workers AI that takes text, images, audio, or video in one request and returns schema-constrained probabilities, built on a 30B mixture-of-experts backbone with about 3B parameters active.
Jev Models & Reimplementations#community#cloudflare#multimodal
Community
The Hugging Face hub for Laya's open System 1 checkpoints: the 421M English model at the repo root, a 322M multilingual build in a subfolder, and the typed-decisions fine-tune, all Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Supersonic Labs' 144.3M-parameter decision model on mmBERT-small: choice, score, and noul answers from supplied options in one pass, 73.15% on 2,000 typed decisions, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
1.3kGitHub stars
A local daemon that pulls open decision models and serves them on TypeSafe's /v1/systemone, so a Jev client switches hosts with TYPESAFE_BASE_URL.
Jev Models & Reimplementations#community#rust#open-models
Community
ollaya-dev/ollayaRust
ollaya-dev/ollayaRustA browser WebGPU export of Julia 1: the same 551 MB decision checkpoint, a Rust WASM tokenizer, and a JavaScript API that answers typed questions client-side with no server or key.
Jev Models & Reimplementations#community#open-models#local
Community
An open distillation of Jev 1.13 onto Qwen3.8-27B: a 108.9M-parameter decision block answers typed questions in one prefill pass and adds a POST /v1/decide route beside a full System 2.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Every decision path has a failure posture, whether you chose one or not. How to put retries in the client, bound the whole call, and make the fallback an explicit, per-action policy.
Jev Practices & Patterns#reliability#retries#api
Read article
The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.
Jev Tools & CLIs#gateway#vercel#cloudflare
Read article
Jev takes text only; Clef-omni adds image, audio, and video input and Liquid d1 adds images, each keeping the same typed-answer shape. Where that matters, where it does not, and the costs the vendors underplay.
Jev Models & Reimplementations#multimodal#open-models#decision-model
Read article
From the Jev community
Posts from builders shipping with Jev right now.