If you only need classification, and you can provide some training data, you can ask Codex/Claude to build an embeddings + logistic classifier model for you

For emails, I get 95% accuracy with this method, with only 50-100 examples for training

Training the model takes less than 5 minutes on a CPU

The resulting model is <1MB, and inference is sub 100ms

Some other cool things about this approach:

* the model doesn’t train on some “ideal” or general classification, instead it learns your preferences

* the model runs on pretty much any mobile device and can be retrained online on the device

* privacy, the whole training and inference is 100% local, no data goes anywhere (except whatever you feed codex/claude while building the model)

Note: to do a more general test, I made a classifier for the Banking77 dataset. The model is <10MB, trains in <30s on CPU and gets 94.5% accuracy, which puts it in the top 5?models by accuracy for that set (the best one is at 94.86%, but it’s 350MB in size and takes hours to train on a GPU).

Man, I'm already burnt out on all this jev talk.

The one thing jev has going for it is a dedicated company focused entirely on making the product good and keeping it maintained. I haven't been willing to jump on board with all these jev-shaped projects because their releases feel driven mostly by opportunism. I'm fine waiting a bit for the opportunists to shake out so we can see who is genuinely committed to bringing something valuable to the open-weight community.

Jev is much better than the traditional ML crowd gives it credit for, but my enthusiasm hits a wall when it comes to their data policy. It is completely draconian. Whatever you feed into the system, they retain.

The jev team needs to release a ZDR product, or their platform is dead on arrival. An open, jev-shaped model will win out solely on that basis.

Found this benchmark for Jev-class models: https://benchmarkheaven.com/jev-models

There are already many Jev-like models in there.

Edit: No affiliation. Just found it and thought others might find it interesting.

If Jev is fundamentally trained using RLCD while you’re building on a Qwen model that was trained using RLHF, how can the resulting model be considered Jev-like?

Bit of a Jev explosion going on. Is it because it's taking us back to a simpler time we understand better? Classification models have been around for a while.

I think a great use case for these will be when they have large context windows and are able to enforce styling rules for frontend development, and component creation rules for react. You can then ditch the styles guides and styling skills and create a decision tree for enforcing styling, so that you can't run into drift issues or duplication issues. That's where I'm wasting most of my time right now, constantly correcting all of the UX/UI issues that are created for every single feature.

I hope these get small and good enough to create “pet like” AIs for games. You know, like scream “follow me” at an NPC, STT stack translates it and feeds it to a local Jev-like model that then picks a number of things for the NPC to do.

Quite impressed by the energy people are putting into making OSS Jev-like models.

I understand the hype but I wonder: what are the use cases for this kind of model? Could it be used in the context of coding agents, or is it more relevant in totally different situations?

Shouldn't Jev-like models be built on top of diffusion models ? like GSAI-ML/iLLaDA-8B-Instruct ?

That showed us the best results at least

I wrote a small proxy that points points to a jev api and a frontier api. My harness connects to the proxy and only sees the frontier api, models, commands, etc. When I send prompts with tool calls, proxy routes to jev, jev narrows the tools, proxy cleans/sends to the frontier api.

So far in my tests, about 60% less tool calls. I'm also going to implement model switching, so it can use cheaper models. I think my workbench harness needs its prompts cleaned up.

I wonder how these would do filtering my spam. I have been using 27B-class models for a while now, and they are nearly perfect at determining what is spam and what isn't. The only disadvantage is computational cost.

On Gemma 4 12B, I am getting 220 ms per move or QS. I used it to play the Snake game locally:

prompt_eval=244 ms wall=245 ms schema_cache=hit generated=0

Move limit reached after 200 moves: score=16, length=19.

So, if a 12B dense model can offer this latency on a local old PC, then definitely you can scale it up with more powerful machines and get even lower latency.

Can someone tell me what is the difference between Jev and a normal neural network that does classification ?

My understanding is: it takes text input and it does one shot classification (no training data)

How are you forcing Qwen to answer in a structured way? I like this one better - https://github.com/deepanwadhwa/OpenDecision

I love that, within days of Jev being released, we have a flurry of open alternatives. This is kinda the open-source dream :)

Because these decision models do not have tool calling, the knowledge cutoff might become a problem. We'll either have to keep training continuously if we run locally or switch to the newer version every month or so when using a closed one like Jev

All the people that are just writing an Jev-like API on top of a normal LLM are missing the point. What makes Jev special is the training data; it's how it's trained. The architecture is probably nothing special. Just a text encoder with parallel prediction branches.

I have tried many of these open-source Jev-like models on some linguistic tasks and they are so bad compared to Jev.

Been hoping for something in this space. Jev-like decision models on Qwen3.5 could really simplify some of our internal routing logic.

Jev feels more and more like a glorified if/else if block

Half-OT:

Is Jev a decoder (e.g., BERT) or is it some kind of encoder (e.g., GPT) that just happens to be trimmed down to only outputting a handful of tokens for the answers and their probability?

- https://github.com/logan-markewich/jeff - https://github.com/bespokelabsai/nimble

Why not name it Qev?

The bright side of Jev being so popular could be that many companies and individuals realize that their applications might work well with a System One model, and they decide to run an open-source (or fine-tuned) version on their own

Using jev for data labeling would be interesting. I wonder how kev compares

How long does it typically take for something like this to become available on openrouter?

How is calibration of Jev or Jev-inspired models being evaluated?

related, vLLM PR nearing merge readiness for Jev like capabilities with diffusion models

https://github.com/vllm-project/vllm/pull/57250

All these Jev projects… great. But Jev was only just released a week ago. That’s the hard limit on how much effort has gone into all these OSS extensions and derivatives: one week. I don’t therefore see any value in adopting any of them, versus just vibe-coding my own if needed.

what kinda of specs would it need to run?

Interesting approach with Qwen3.5 for decision models. Curious how "tiny" they've made them while keeping LLM reliability for critical paths.

OK - I guess I'll ask here.

As there have been a lot of Jev related submissions, can someone point me to a simple guide on how I can use it? For example, say I have a script/workflow where I use OpenRouter for LLM calls, and at some point I want to do a simple classification. Can I still use OpenRouter with some Jev model...?

Interesting to see a Jev-like approach applied to Qwen3.5. Always appreciated Jev's simplicity for quick decisions.

Distilling Jev should be super easy and cheap.

looks high lev

Oh Jared is cool - he made After and Razzle - nice

[dead]