Kev: Tiny Jev-like family of decision models built on top of Qwen3.5
- ID
- 26905
- Status
- summarized
- Published
- 21 Sep 2026, 3:11 PM
- Fetched
- 23 Sep 2026, 3:29 PM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://github.com/jaredpalmer/kev/tree/main
- Source URL
- https://hnrss.org/best
Summary
- Score
- 7.0
- Created
- 23 Sep 2026, 4:38 PM
- Tags
- Audience
- developersai_agent_usersai_ml_learners
What happened
Kev is a family of small decision models (0.8B, 4B, 9B) built on Qwen3.5 that handle yes/no, multiple-choice, and rating questions in a single request, with questions sharing input text but unable to read each other. The 4B and 9B models fit on a 32GB Mac using bf16, and the API matches TypeSafe's System One Python SDK so you can point existing clients at a local server. Pretrained weights, training code, and eval data are all included.
Why it matters
If you're building agent routing, triage, or classification pipelines and want to avoid per-call API costs, Kev lets you run a purpose-built decision model locally on a Mac or CUDA box — the 4B model is a one-command `uv run` away. The System One API compatibility means you can prototype against their hosted SDK and swap in your own server without rewriting client code.
Discussion angle
Compare Kev's approach — tiny purpose-built decision models that only output yes/no/choice/score — against using a general-purpose LLM with structured output for routing and triage. Where does the smaller model win on cost and latency, and where does it break?