README

Kev

Small Jev-like decision models you can train and run yourself.


  
  
  
  
  


Kev is a family of small decision models built on Qwen3.5 and Qwen3.8 and based on the architecture described in Jev's Architecture Unmasked. You can use the pretrained weights or train your own. The API matches TypeSafe's System One, so you can point their Python SDK at your local server.

 Highlights

- Yes/no (noul), multiple-choice (choice) and rating (score) questions in one request. The questions share the text but can't read each other.
- Calibrated probabilities by default: each checkpoint ships with a fitted temperature.
- Drop-in for Jev: the TypeSafe Python SDK works against a Kev server unchanged.
- Four sizes, versioned together as Kev 1.0: from a 0.8B that runs on a laptop to a 27B for a single data-centre GPU.
- Documents of up to 65,536 tokens, on CUDA and on Apple Silicon through MLX. Each model card says how long a document can get before accuracy drops.
- Fine-tune on your own labelled examples. A coding-agent skill runs the whole loop on Modal, from finding your questions to serving the result.
- Deploy your own HTTPS endpoint with one command. It scales to zero when idle.
- Try it in the browser first: huggingface.co/spaces/jaredpalmer/kev.

 Models

Start with Kev-4B. Move to Kev-9B if you have a bigger GPU, or to Kev-27B if you have an 80 GB GPU and want the most accurate Kev. Use Kev-0.8B when size matters more than accuracy.

| Model | Base (license) | Runs on: CUDA | Runs on: Mac (MLX) | Validated context | Held-out datasets: index | Card |
|---|---|---|---|---|---|---|
| Kev-0.8B | Qwen3.5-0.8B-Base (Apache-2.0) | L4, any 4 GB GPU | Any Apple Silicon Mac; measured to 65k tokens | 8,192 | 23.3 | Details |
| Kev-4B | Qwen3.5-4B-Base (Apache-2.0) | L40S, H100 | 32 GB Mac; measured to 65k tokens | 8,192 | 38.0 | Details |
| Kev-9B | Qwen3.5-9B-Base (Apache-2.0) | L40S, H100 | 32 GB Mac or larger (expected, not measured) | 8,192 | 41.0 | Details |
| Kev-27B | Qwen3.8-27B, post-trained (Apache-2.0) | B200, H200, H100 80 GB | 96–128 GB Mac (expected, not measured) | 65,536 | 52.3 | Details |
| Jev | Hosted | TypeSafe's API | – | – | 54.0 | – |

"Held-out datasets" is the chance-corrected index of the community Decision Index, scored on the test split of breadth-v1: 14 public datasets in five areas that no Kev trained on. "Validated context" is the longest document, in tokens, for which accuracy on real contracts (CUAD) stays within 3 points of the same model's accuracy at 8k tokens, at the 95 % lower bound; each model card has the measurement by length.

| Model | Accuracy: New Sources | Accuracy: Trained Sources | Brier: New Sources |
|---|---|---|---|
| Kev-0.8B | 0.648 / 0.697 | 0.827 / 0.838 | 0.481 / 0.416 |
| Kev-4B | 0.817 / 0.838 | 0.873 / 0.865 | 0.269 / 0.242 |
| Kev-9B | 0.820 / 0.852 | 0.874 / 0.873 | 0.289 / 0.217 |
| Kev-27B | 0.851 / 0.889 | 0.865 / 0.866 | 0.225 / 0.156 |
| Jev | 0.857 / – | 0.845 / – | 0.211 / – |

Each cell

Top issues

Latest release

Kev 1.0 is the first versioned release of the whole Kev family: four decision models that read a document and a set of typed questions and return calibrated probabilities over the options, in one forward pass, behind TypeSafe's System One API. Nothing in it is newly trained. It pins the checkpoints, model cards, evaluation suites and serving code that the next generation of Kev is measured against, with the same tag (v1.0) on every Hub repo.

 What's in 1.0

| Model | Hub repo | Weights revision | Form | Base | Temperature | Validated context |
|---|---|---|---|---|---|---|
| Kev-0.8B | jaredp