Jev vs Kev
Jev is TypeSafe’s hosted System One model. Kev is an independent Apache-2.0 family of Jev-like models you train and serve, matching System One HTTP shape.
This unofficial note restates what each project publishes. It is not a bake-off and not an endorsement.
What each project is
Jev is TypeSafe’s hosted System One model. Official usage is the TypeSafe API (POST https://api.typesafe.ai/v1/systemone, model jev-latest) or a TypeSafe SDK. The Python client reads TYPESAFE_API_KEY.
Kev is published at jaredpalmer/kev as a family of small decision models built on Qwen3.5 and Qwen3.8. The README states the API matches TypeSafe’s System One, so the TypeSafe Python SDK can be pointed at a local Kev server. License is Apache-2.0. A browser trial is documented at huggingface.co/spaces/jaredpalmer/kev.
Kev is not a TypeSafe product and does not ship TypeSafe weights.
Documented question types
Both projects document the same three question names in one request:
| Type | TypeSafe Jev | Kev README |
|---|---|---|
| Choice | Closed set → choice, probabilities, confidence | Multiple-choice (choice) |
| Score | Ordered rubric → score, probabilities, confidence | Rating (score) |
| Noul | Yes/no → noul in 0–1 | Yes/no (noul) |
Kev’s README says questions share the text but cannot read each other. TypeSafe documents the same isolation: questions evaluate in parallel against the same state.
How you run Kev locally (project docs)
The README asks for Python 3.12 or 3.13 and uv. Torch is documented as having no wheels for 3.14 yet.
git clone https://github.com/jaredpalmer/kev.git && cd kev
uv sync --extra serve
uv run --extra serve python -m kev.serve --run jaredpalmer/kev-4b --port 8009The first run downloads the adapter and base model. The serve path is documented to use CUDA or ROCm if a GPU is present, or MLX on Apple Silicon.
Example SDK pointing from the same README (placeholder key only):
from typesafe_sdk import TypeSafeClient
client = TypeSafeClient(
api_key="local",
base_url="http://127.0.0.1:8009",
model="kev-latest",
)Sizes the Kev README lists
| Model | Base (README) | Documented “runs on” |
|---|---|---|
| Kev-0.8B | Qwen3.5-0.8B-Base | Any Apple Silicon Mac, L4 |
| Kev-4B | Qwen3.5-4B-Base | 32 GB Mac, L40S, H100 |
| Kev-9B | Qwen3.5-9B-Base | 32 GB Mac, L40S, H100 |
| Kev-27B | Qwen3.8-27B (post-trained) | B200, H200, H100 80 GB |
The README names Kev-4B as the starting checkpoint. Accuracy tables on that page are author-published on the authors’ splits; this site does not repeat those numbers as an independent ranking.