yesterday at 9:59 PM
Forgive my lack of understanding but how long before Jev type functionality is just built straight into all frontier models?
today at 1:06 AM
Do people really have zero awareness that Structured Outputs with a constrained schema has been a thing for a while now, and open weight models that give you logprobs can give you distributions per key?
Like what am I missing? I use Structured Outputs every day and this just seems like that with fewer steps?
edit: Where I'm coming from, I can triage 10,000 support tickets with deepseek flash for less than $1, and latency is sub 1 second if it needs to be integrated into a live user flow. I don't need anything cheaper or faster than that.
today at 1:31 AM
FWIW you’re absolutely correct on how most developers are unaware of this as they’re mostly operating on the API level and not aware of server side (vLLM/SGLang) capabilities.
The aspect that I like the most is the typesafe API that introduces new probabilistic concepts that are more sound than json schema and constrained decoding with quasi-confidence scores. Developers were asking LLMs to also emit confidences which made absolutely no sense whatsoever.
today at 1:35 AM
Can you link to some info on this? I am just starting to poke around in the space and would love a leg up.
today at 1:54 AM
This is an example implementation that was making the rounds recently. If you point whatever model you prefer at it, it’ll do a good job of explaining it.
No comment on this model itself, might be over fitted to the Jev benchmarks just to beat it.
https://github.com/Mushroom-Systems/lichen
today at 1:42 AM
I don’t know of one definitive resource but “constrained decoding”, regex/grammar-based decoding, json-schema decoding will give you a bunch of hits. Look up vLLM, SGLang and outlines’ implementations for more technical details.
This looks pretty decent: https://www.aidancooper.co.uk/constrained-decoding/
today at 2:11 AM
I remember reading a paper about a model that used a parser bound to the end of a LLM where it squashed all logprobs for outputs that were parsing errors. I have not seen this used in the manner I thought I would. I thought it would have been entirely possible for a model to construct it's own grammar for how it would prefer to respond and then opt to generate tokens that matched (with a metacode to turn it off obviously).
That said, I think the advantage of Jev style approaches is not their capabilities, but rather the capabilities that they have for a much lower resource requirement.
today at 1:13 AM
> Like what am I missing?
Orders of magnitude faster and cheaper answers.
Sure a MacBook Pro can control a servo motor but maybe an arduino or a cheaper microcontroller for deployment?
today at 2:19 AM
Excellent analogy
today at 1:21 AM
Many people yes, but it’s probably for the best. Without fine-tuning on such style the results would be unreliable. It wouldn’t be the probability of the outcome of whatever you intend, but just the next token probability - which could alter if you add a space or punctuation to the prompt. Very flaky.
today at 1:08 AM
jev is optimized for it, a standard LLM isn't and it also costlier and slower.
today at 1:21 AM
Pricing
The same reason I want diffusion language models to be mainstream
today at 1:08 AM
You don't even need a constrained decoding.
As a closed source chat-API provider, you just need to find a way speak JSON correctly at API output.
today at 1:08 AM
Price and speed are the difference. But deepseek flash is usually fast and cheap enough, so the use cases are somewhat limited.
yesterday at 10:08 PM
If I had to guess, it's already built and is just waiting on Product's/Marketing's desk. How do you position this without looking like your roadmap is being determined by newcomers? Probably don't want to adopt the same verbiage+acronyms - but also can't be seen to be just sherlocking features.
yesterday at 10:17 PM
I think Apple has demonstrated that shipping second has essentially no negative impact if your product is seen as higher quality.
yesterday at 10:22 PM
I think Nvidia has demonstrated that shipping first is a multi-trillion dollar opportunity if you don't shy away from a challenge.
yesterday at 10:30 PM
I think it depends on the model: Nvidia ship products with APIs, apple/jev/etc. ship end user products. The former is much more subject to lock-in, increasing the value of early market adoption because there's a 3P Nvidia ecosystem sprung up in response. Apple/jev/etc. do have APIs & corresponding 3P ecosystems but those are usually a smaller component of market capture than direct product end users, so the space ends up more competitive.
yesterday at 11:26 PM
NVIDIA literally hired all of the 3DFX team and patents after they already proved the graphics accelleration hardware market was massive with the Voodoo card line
In fact NVIDIA wasn't even a close competitor to 3DFX in the graphics card game at that point
today at 12:07 AM
It panned out great. 3DFX filed for bankruptcy less than 18 months later, and Nvidia could pivot from designing raster chips to considering CUDA's architecture.
It's not like 3DFX was the first GPU vendor. Nvidia saw the opportunity to be the first true GPGPU vendor, and they beat their competitors.
today at 12:55 AM
IBM shipped the first PC GPU (the Image Adapter/A, 1989) following on the first PC 2D accelerator (the 8514/A, 1987). There was zero benefit to being first. 3dfx had the first mass market, cheap GPU in 1996.
ATI (AMD) in 1989 copied the unpatentable parts of the 8514/A, improved it, and went on to dominate 2D accelerators.
nVidia’s first 3D card was a complete flop. They did not achieve success for years.
Nvidia is in the right place at the right time.
today at 2:02 AM
not to be _that_ guy, but S3 was the dominant force in 3D. ATI was nowhere to be seen with their Mach/Rage crap. It was an S3 era, and then seemingly out of nowhere 3dfx swept in with Voodoo (not seemingly though - 3dfx came out from imploding SGI). Only a bit later Nvidia, after it recovered from NV1 fiasco, brute forced relentlessly from Riva 128 to TNT to TNT2 to Geforce 256 which broke 3dfx (along with their lack of business sense) and had Nvidia bought them. ATI did their 2D schtick only during that timeline while S3 stumbled with Virge and ATI didn't come as a threat until Radeon; After they bought ArtX (which did Gamecube graphics) which turned into Radeon. S3 died off in that race with Savage3D and the only other major players were Matrox and 3dlabs. Matrox kind of found temp refuge in video segment, and 3dlabs infamously pushed for OpenGL 2 and survived for a bit on 3d workstations. The most surprising (IMO) was the downfall of E&S which kickstarted most of the things mentioned. E&S and Real3D are both a great story in themselves how first movers can become absolutely forgotten and obscure real quick (with Intel740 ended up in ATI).
today at 1:43 AM
We're talking about GPGPU products, not raster GPUs. I cleared that up pretty well in my last comment.
yesterday at 10:49 PM
today at 12:29 AM
It's more due to ineptitude of competition. If AMD shipped second, but better product it would be another thing but it is still a bit of a mess of an ecosystem on AMD side
today at 1:46 AM
It's not AMD's responsibility to dethrone Nvidia any more than it is Apple's. AMD sells CDNA, but the momentum is with CUDA and Khronos can't get anyone to sit at the same table anymore.
yesterday at 10:33 PM
Chip manufacturing intrinsically comes with one hell of a moat. There's not much parallel in software.
yesterday at 10:54 PM
That’s kind of funny because Nvidia’s biggest moat is arguably CUDA, the software ecosystem around their chips
yesterday at 11:30 PM
Historically there was a big patent moat in (Graphics) GPU design. This continues with CUDA, but obviously Intel and AMD could find ways to support eg PyTorch. What we don't know is how much effort that cost them, or why they decided they couldn't make a CUDA API compatible competitor.
today at 12:51 AM
Intel could have beat the pants off Nvidia a long time ago with Arc if they’d bothered to ship usable drivers. But they refuse to, and simply can’t figure it out, so Arc cards remain cheap because they’re so #%£€ing hard to get working well, and everyone is nervous they’ll lay off the driver team again.
today at 12:54 AM
AMD has always had software problems too. Hardware companies often devalue software and suck at it.
If you can make them work Arc cards are a massive bargain. On raw compute the silicon is not bad.
yesterday at 11:13 PM
CUDA is a lock-in moat, the infrastructure needed for chip manufacturing is a barrier-to-entry moat. Two different things.
yesterday at 11:34 PM
There are many microarchitecture patents used in NVIDIA chips. I'm sure they have a team that rips apart AMD chips looking for infringement. The way CUDA works is tied to many GPU architecture decisions and it would be hard to decouple them efficiently. Obviously a huge effort was made to get PyTorch decoupled from CUDA.
today at 1:02 AM
I’m totally guessing, but I can’t imagine CUDA has any significance at the frontier lab scale. The operational and capex costs are so massive that the convenience of the platform becomes a minuscule consideration.
It’s just that Nvidia’s stuff works, and available at scale, and includes the full stack with networking, cooling and such.
today at 12:03 AM
If cuda is an API, and llms make apis effortless, then how big of a moat is cuda?
today at 12:52 AM
IMO, yes a lot of the "nvidia pays lots of people to make non-portable, tightly coupled backends to open source projects" is potentially going to be less of a moat?
(Though it could turn into "nvidia pay lots of people to use LLMs to make non-portable, tightly coupled backends to _even more_ open source projects")
today at 1:10 AM
It turns out the moat is “writing drivers that work”; Nvidia drivers simply work, and Intel’s are poor quality. So if I want stuff that works I need to buy Nvidia gear.
today at 1:40 AM
There is zero stigma to shipping second. If anything, the labs’ customers are probably begging for them to add these features natively so they don’t have to deal with the hassle of adding another provider to their stack.
today at 12:26 AM
Negative three years, give or take. Although recent Anthropic and OpenAI models no longer expose the capability. But for any open model you just tell it to respond with a single token "Y/N" and take the logit difference. If you want multiple distinct questions answered you just ask them independently and put the shared context first so it gets cached.
OpenAI and Anthropic don't want to give out logprobs these days but could trivially add a dedicated classification API to their existing models if there was enough demand.
today at 12:32 AM
I think the main differentiator offered by Jev is not the ability to answer questions, most models can be coerced into that function if they don’t already have a dedicated pipeline for it, rather it’s the extreme speed of the evaluation, and very low cost that opens new possibilities.
today at 12:55 AM
> it’s the extreme speed of the evaluation, and very low cost that opens new possibilities.
It also returns confidence scores for all choices.
Granted, they are not stable. They fluctuate even when you reorder choices, but it still counts as an additional feature.
today at 1:24 AM
They’re useable if you’re trying to rank autosuggestions or something, not great at implying actual understanding
today at 12:34 AM
The evaluation is fast because it's all prefill computation with only a single token of inference. Ditto cost, you're paying 100% input costs and nearly zero output. There really isn't any architectural magic to Jev, it's just a straightforward application of normal LLM tech with some good marketing.
I mean, Jev is also probably cheaper because it's a rather small model (or at least, I suspect it is based on the overall level of intelligence it demonstrates) so that helps make it cheap too.
today at 12:42 AM
Yes this is exactly my assumption as well. I think ppl forgot before agents it was expected slo to have a ttft in range of a few hundard ms, which is what jev is also achieving.
Thats also why it feels weird they say they don't "charge for output tokens" since its literally generating a single (or at most very few tokens).
yesterday at 11:00 PM
Probably at the frontier stage - you will only see it where Jev is better regardless of cost.
For everyone else who is conscious of cost, you're already seeing this being built into harnesses.
Almost certainly, you'll see versions of this from all the Chinese labs as fast as humanly possible.
If I had to guess, Cursor/Grok or Google/Antigravity will be the first major players to natively support something like this to drive down cost, as they're primarily the budget conscious choices.
I would be astounded if Anthropic leads the way on a cost reduction.
yesterday at 10:02 PM
No need, as that functionality can run locally no problem.
yesterday at 10:51 PM
the appeal would be if they can deliver it at a much higher performance and similar speed, which is plausible
yesterday at 10:06 PM
BeRT and FLAN-T5 were used as classifiers 5-7 years ago, they were technically "frontier" for their time.
yesterday at 10:17 PM
This is the exact comment I’ve been waiting for, what is the difference between classifiers and jev?
yesterday at 10:32 PM
Jev is a classifier. The big thing about it is that it has high accuracy on domains it wasn't fine-tuned for, like an LLM, but with speed and cost comparable to traditional classifiers.
yesterday at 10:37 PM
BERT need to be fine tuned for your use case, Jev generalizes. It’s a pretty big difference!
today at 1:47 AM
The fact that its not narrow and stupid is what's different
yesterday at 10:57 PM
FLAN-T5 generated text (Jev does not generate text), and BERT wasn't able to do tasks without fine-tuning.
Jev is basically a kind of FLAN-BERT, if you want, where it has built-in multi-task ability, but doesn't generate text. It only generates 255 floats all at once, making it much faster, and what those floats mean (if anything) depends on the prompt.
Eg, the following query is put in the encoder model:
{"question": "Rank these 5 things by increasing order of how big they are", "choices": ["truck", "cow", "mouse", "ant", "building"] }
The model returns [3., 2., 1., 0., 4.], and 249 other meaningless floats that are hidden from you by the UI.
The UI stitches the first 5 floats with the choices and returns something like:
{"rank": ["ant", "mouse", "cow", "truck", "tower"]}
today at 12:30 AM
How does it know that you’re asking for “rank” instead of something else if it’s not generating text?
today at 1:16 AM
By reading, not generating, text.
yesterday at 11:37 PM
So it generates logits in a 255 token output space? ;)
today at 12:44 AM
logits assumed some form of softmax or logistic, which may not be the case
yesterday at 10:20 PM
nothing in utility. we used various bert variants to satisfy our usecases and still in uses. free and they run locally.
yesterday at 10:27 PM
yesterday at 11:40 PM
But for those of us that prefer open source and self hosting, JEV alternative LAYA will beat anything the frontier models package up.
yesterday at 10:53 PM
I don't think there would be any utility for that. Anything jev can do, a frontier model can also do. Just not as quickly or as cheaply.
yesterday at 11:06 PM
I think these products (Jev and the inevitable offerings from Anthropic, OpenAI, etc) want to become more than end-user output machines. They'd benefit from being in the hotpath of other services. Not backgrounded generation but in-band, request-time work.
1M x $0.50 == 1B x $0.0005
yesterday at 11:22 PM
To expand a bit for my current use cases. Inline routing of work to heavy task specific models, and prompt/context generation (user is asking something, what and how much should we prompt the expensive LLM with). Latency or time to first token does matter for some applications.
yesterday at 9:45 PM
I compared it to Jev in my current use cases and it's very inaccurate. 70% vs 94% . for classification, it's unacceptable.
yesterday at 10:05 PM
For _your_ classification it's unacceptable. The OP seems to have anticipated this and mentions you can fine tune it for your use case. Did you try that?
I don't think the point is to displace Jev, but to show it's possible to build an MVP on open weights without years of work and millions of dollars.
Why (presumably) an engineer would dismiss exploring a lightweight, custom alternative to locking into a fashionable PaaS, I'll never know.
yesterday at 10:53 PM
> The OP seems to have anticipated this and mentions you can fine tune it for your use case. Did you try that?
You can already so that with classification models such as ModernBERT, at 0.4B.
Jev's value is its zero shot performance without having to fine-tune.
today at 12:20 AM
I am sure I am missing something obvious here, but why is that valuable? Like, what kinds of projects are there where you need to classify stuff but are unable to make a bespoke model targeting the specific problem?
today at 12:37 AM
For my org, it meant we could trial classifiers across various internal systems with little to no engineering effort. In one case we ended up building our own classifier instead of Jev, but in others we kept Jev because it was zero-effort for a great impact.
today at 1:19 AM
> unable to make a bespoke model targeting the specific problem?
There is a fixed cost (and some maintenance) to e.g. fine tuning ModernBERT.
Maybe once you include all of that it might be a half-day to a day of engineering time to set everything up in a maintainable fashion.
For Jev, it takes all of 30 seconds of prompting. And it's not that much more expensive to deploy vs. a BERT model.
yesterday at 10:17 PM
Not sure the task at hand here. But if it doesn’t require any reasoning/thinking and it’s just a classification task, it’s worth a shot to look into training your own classifier
I’ve run some benchmarks. Using embeddings + logistic classifier, the architecture matches or beats Jev and Laya in all basic classification tasks (datasets tested: AG News, Emotion, MASSIVE Intent, Banking77) The type of task in which it does really well, especially against Laya, is classification with >50 classes
The classifiers also run in <1ms, so they can be very fast and precise at the same time
But this architecture has no “reasoning”, so it performs rather poorly on tasks that require it, like the ones from the XLNI dataset (Jev/Laya do a lot better on this one)
For the latter cases, you could use add a local lightweight LLM, something like a Gemma model. Or even some basic MLP, depending on the tasks/data
yesterday at 11:54 PM
What do you use to determine that a particular task in a heterogeneous pile of tasks requires reasoning? The logistic classifier itself is too dumb to recognize the details of the problem that make it reasoning-sensitive (IIRC recognizing the “fiddliness” of a given problem requires a recognizer at least as complex as the problem itself.) And if you’re using the lightweight LLM for that, then you may as well skip the classifier and just use the LLM all the time, since that eval step is already going to be dominating your response time anyway.
My understanding of Jev is that it’s a replacement for the LLM you’d necessarily need to use to identify reasoning-sensitive workloads in a heterogeneous mix, where Jev will be cheaper than an actual LLM and so act as an actual optimization / de-bottlenecking change.
today at 12:24 AM
I’m in the process of piecing together the different task/dataset-specific classifiers
Depending on how much overfit, you can go from routing deterministically based on features/shape of the input data, all the way to training a routing model (which could be a classifier too). I’ll need to experiment to find the best approach
For completely unseen/unexpected, I’ve also experimented routing to a local LLM: request comes in, if there’s a marching classifier, send it there, otherwise send to LLM+training. As the system learns more tasks, the % of requests that go to the LLM go down over time
today at 1:20 AM
I have not personally reviewed the benchmarks, but recall seeing a post where a Linear SVM with bag-of-words features outperformed Jev on many simple NLP classification tasks, like the ones you cite. You don't even need embeddings!
yesterday at 11:10 PM
"Using embeddings + logistic classifier, the architecture matches or beats Jev and Laya in all basic classification"
Have I understood correctly that you trained only the logistic classifier, but didn't need to train the embedding model?
If so, I'm curious whether you compared that approach (A) with:
B) Jev only, with a single output.
C) Jev with multiple outputs fed into a logistic classifier.
Obviously C has cons (can't be self-hosted, needs some up-front work on deciding the shape of the output) but it might be somewhat more interpretable. (And I suppose it might have better performance?)
yesterday at 11:46 PM
You are correct, I didn’t train the embeddings model
Here's a gist with code you can use to test the Banking77 dataset: https://gist.github.com/nicobrenner/056a5aaff5d0119c0032ecda...
The gist uses BAAI/bge-large-en-v1.5, which is 1.2GB approx. You can replace it for all-MiniLM-L6-v2 (91 MB @ fp32 or 45 MB quantized fp16) small enough for mobile/edge. With all-MiniLM-L6-v2 it still gets 93.0% on Banking77, only 1.3 points behind bge-large at 15x smaller
I haven’t compared different ways of sending requests to Jev
The data to train the classifiers comes from the datasets used to test them (not from Jev)
yesterday at 10:13 PM
Needing fine-tuning for the use-case completely changes the product category
today at 12:26 AM
[dead]
yesterday at 11:18 PM
My use case is simple classification for job ads. Things like, industry, work settings (remote, hybrid, onsite) and job type (full time, part time .. etc).
I did side by side comparison with Gemini 2.5 Flash Lite, Jev, Jeff
I tried the 0.8B model, completely useless in classification. Qwen Jeff-Qwen3.5-2B was better, but still missed job type.
I suppose with larger model, this could be useful, but would require more ram and will be slower.
today at 1:56 AM
i've tried all the "open source" me too Jevs
they all suck
yesterday at 10:37 PM
What proportion of commercial LLM use is classification? I'm just wondering what happens to business AI spending/data centre usage when they realise they don't need full LLMs.
today at 1:19 AM
I'd expect the vast majority at this point is coding. Classification is a thing but in my experience tends to run on light, cheap models, not the proprietary frontier ones.
today at 1:07 AM
The American stock market probably loses 20% of its value in a few days.
today at 1:24 AM
For those of a certain age - the fact that Askjev.com is still available astounds me.
today at 1:29 AM
I can’t believe ask.com threw in the towel in 2026. Like ride that LLM hype - how much better could the SEO for ask.com be? Do a pivot!
today at 1:27 AM
not anymore
today at 1:28 AM
Awesome, I was just looking for a decision-making model that can be deployed locally, and here you are. Thanks a lot!
yesterday at 10:16 PM
Typesafe has been quite about the underlying technology behind Jev. Given the speed and cost my hypothesis is that it doesn’t input tokens the way that LLMs do, ie iterating over every word and drawing the connections between each. That is an o(n^2) problem which is why LLMs are so expensive as they scale.
yesterday at 10:22 PM
Most likely: it does still have attention layers (the O(n^2) part), but it’s not autoregressive (which makes it O(n^3) because you have to run the whole model again for each predicted token)
yesterday at 10:58 PM
I feel like the most likely is that it largely works like how all the recent copycats work: take an existing llm, modify the decoder, do RL training. I think the main reason that jev works better is that they spent more time on that post training step.
today at 12:01 AM
My guess as well
yesterday at 10:42 PM
Then again, it's only a very small number and fixed set of tokens for the output.
today at 12:36 AM
The main difficulty with fast (low-latency) inference, is not actually computation but loading parameters from memory. The problem with generating 1 token at a time isn't that that its expensive computationally (it is, but so is training), but that you need to stream your entire model from memory for every single token (and also the KV cache but that's besides the point). So the strategy usually taken is to process big batches of user requests, letting you share the memory loads across users. This means individual answers aren't that fast, but you can do lots at once. This is why local inference isn't cost-effective; its because the model weren't designed for it in the first place.
My hypothesis for Jev is that they simply generate many answers independently in parallel from your prompt, and then discard the duplicates (or train to avoid duplication with attention between the ). In that way the entire batch is 1 user's prompt.
today at 1:04 AM
My guess is that they use an encoder-only model as the foundation and then do RLDC. Why? 1) it doesn’t need text generation 2) limited context window (40k last time I check)
Those are telltales of a Bert model.
yesterday at 9:38 PM
This type of project looks extremely useful. There was a lot of buzz around Jev, but having models that run locally and can be fine-tuned is extremely helpful.
today at 1:49 AM
I think Jev is still the king and while I really appreciate this and other similar projects, Jev gives assurance and value that is hard to beat. Well.. this works locally which is always best, even if slower.
yesterday at 11:55 PM
Isn't jev just a less nuanced classifier? What am I missing?
today at 12:11 AM
The zero-shot learning is what makes these different from a traditional classifier.
today at 12:25 AM
Zero-shot just means you give it zero examples. Jev lets you add examples, so it's one-shot/few-shot depending on how many you provide. After playing with it, it seems like once you wonder off their guide examples domains - you have to provide examples to get anything useful from it.
IMO you going to get better results by doing a small tune of a tiny model. Making training dataset for it with LLMs is easy, serving it is going to be cheaper.
yesterday at 9:59 PM
Von 1.2 had a better Doom score :D
yesterday at 10:10 PM
Oh wow, those Doom scores for Jeff are pretty terrible
The Von numbers have led me on a rabbit whole of getting a classifier to play Doom
I got it to average 22 kills (the max is 26) on that same scenario that Jeff and Von are testing on (it’s called Defend Center)
Now I’m having it play a more advanced scenario, and it’s doing about 45 kills (SOTA is ~59 kills)
It’s amazing what you can do with small classifiers if you can collect some data. These models I’m testing train on CPU in seconds (what takes the longest is running the game, doing test runs and collecting data), they are <1MB in size and do inference in <1ms on CPU
Edit: after looking at Jeff's numbers more in detail, the 6.5 kills number is not that bad, but it can definitely be better ;)
yesterday at 10:51 PM
One day we’ll be using the same Kills/SOTA metric for models driving physical kill-bots.
yesterday at 11:16 PM
Anything like this in the VLM side? Classification on images...
today at 12:04 AM
I've always been more afraid of these types of models than LLMs. These are what enable mass surveillance at scale and autonomous real time combat drones. Now they are spreading and being optimized. Gg.
yesterday at 10:15 PM
Can we get a price comparisson ?
Edit: Running them for the masses.
yesterday at 10:59 PM
I assume the cost is whatever you run the model on. Jev is already dirty cheap, $0.42 per million input tokens and output is free and the input cost is covering tokenization + API utilization
yesterday at 11:07 PM
Well, I'm no expert, but this runs on Qwen3.5 and Gemma 4, stripped down, but they're pretty pricey.
yesterday at 11:18 PM
$0.042 not $0.42
yesterday at 11:08 PM
My name jeff
today at 1:46 AM
No, he jefe, man.
I've been holding it in since I saw the title. I was surprised it wasn't all over repo!
today at 1:18 AM
[dead]
yesterday at 11:12 PM
[dead]
yesterday at 10:04 PM
[dead]
yesterday at 8:23 PM
Hi HN. Jeff is a set of small, open-weight Qwen3.5 and Gemma fine-tunes for zero-shot classification, with respectable out-of-the-box performance, meant to be slotted right into code (or fine-tuned further as needed). You give them a situation and a list of options; they return a calibrated probability for each, in one forward pass, with no text generation. The 2B scores 83.1% on a five-benchmark panel (Jev's published figure: 83.0%); the 0.8B decides in about 28 ms on an M4 Max. Apache 2.0, with a Jev-compatible API (I'm not affiliated with TypeSafe).
When TypeSafe released Jev a couple of weeks ago and then AutoJev appeared, I wanted to see if I could replicate the experiment using only small language models on local hardware. So everything ran at home: one RTX PRO 6000 for training, two DGX Sparks running Qwen3.8-Flash-Next to write the synthetic data, a MacBook for testing, all monitored from my phone over Tailscale.
The caveat: the published Jev and AutoJev numbers are on a different sample of the same benchmarks, and Jeff's overall score comes from classification-style tasks (96% on Financial PhraseBank, 86-89% on RAGTruth, both above Jev). On multi-step reasoning it's behind: BBH 64-68% against Jev's 94%, and about 50% on JevBench's hard tier against 73%. That isn't surprising, and I don't think it matters: no 0.8B or 2B model reasons like a large one, and nobody should expect it to. These are extremely fast judgement-callers. In one of my apps I use the 0.8B for voice navigation; a quick fine-tune (about half an hour on one GPU) took it from 32% to 96% on held-out commands, at about 40 ms per decision.
The fun part: games, as a zero-shot test. Games aren't the ideal zero-shot test, but they're fun, and TypeSafe did it with Jev too. There was no game data in training. Each turn the code describes the situation and the moves in words, and the model picks one; the options say what each move leads to, never which one is right. Over 20 episodes each:
- Doom (ViZDoom): Jeff 0.8B 6.55 kills per episode, the same as a hand-coded bot and as Jev's published run. Jev's prompt spells out the aiming rule and takes about 212 ms per call over its API; Jeff gets "the nearest monster is a little to your left" and decides in about 29 ms on my Mac.
- Frogger: 10.3 crossings, level with the hand-coded bot (10.25), and 10x the untrained base model (1.0).
- Pac-Man: 57 of 98 pellets, about 60% of the bot's score and 2x the untrained model.
Videos of every run are linked in the README.
Lessons learned:
- System 1 models are here to stay. Being able to process unstructured data at software speed inside an app is extremely powerful, and being able to do it locally is fantastic.
- A small model is a classifier, not a planner. Models of 0.8B-2B don't reason like Qwen3.8-27B or Jev, and they don't need to: present the options the right way and you get 40+ decisions per second, depending on your hardware.
- Fine-tune it if needed. If zero-shot isn't good enough for your task, a short fine-tune on your own examples is.
- Wording matters enormously. Giving Frogger's final step the same words as every other forward option ("safe, and one row closer to the goal") took one episode from 15 crossings to 23. Before that, the frog just stayed on the last log.
- Bigger isn't better. The untrained 2B is already more risk-averse than the untrained 0.8B (in Doom it prefers turning away from the nearest monster), and training made it hesitate in Pac-Man. That's probably why the 0.8B beat the 2B.
- Benchmarks don't predict play. Untrained Gemma 4 E2B beats both untrained Qwens on the benchmarks (62.5%) and plays every game worst: right most of the time, but not reliably, and in a real-time loop the mistakes compound.
yesterday at 9:09 PM
Funny that the 2B loses to the 0.8B. Question about the benchmarks: BBH and JudgeBench are more reasoning, where you fall behind, but for zero-shot classification there are more relevant ones like Banking77 or CLINC150. Was there no temptation to pick something closer to where System 1 models are actually used?
yesterday at 11:55 PM
Great project! How does this compare to asking qwen to reply with 1 token in terms of speed?
yesterday at 9:56 PM
That's some local hardware.
yesterday at 10:22 PM
Was going to make the same observation. Cool to run locally - but renting seems the saner choice?
Would love to see what this run would cost from something like Verda. Just out of curiosity. I'm not going to be installing any sparks at home anytime soon.
yesterday at 8:54 PM
[flagged]
yesterday at 9:11 PM
First commit of six commits was 4 hours ago.
yesterday at 8:57 PM
I'm curious: in the age of AI, where you can literally tell your clankers to work on something, why do you even care about the longevity of open source projects? There are several projects that are abandoned, and I have resurrected them for my own use cases without any issues.
yesterday at 9:02 PM
By the same reasoning, why should I care about the output of someone else’s clanker if I can just get it from my own clanker?
But to answer your question, I do still care about software being maintained by someone who can make design decisions instead of just yielding control to bots who tend to produce mediocre designs.
yesterday at 11:10 PM
These are all good questions.
Perhaps we can think of projects like these the same way we’d think of a good tweet - not valuable in itself but valuable because of the ideas or conversations it generates.
yesterday at 9:18 PM
Maybe because they know more about the domain than you do, and it will take you a lot more back-and-forth with your clanker than it would for them with their clanker.