jaredpalmer
Jared Palmer: Kev 4B
jaredpalmer/kev-4b
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Kev-4B is a small open-weight English-language decision model by Jared Palmer: a LoRA adapter and pointer head on a frozen Qwen3.5-4B-Base that scores user-supplied options instead of generating text. Given a document plus typed choice, score, or yes/no questions, it returns calibrated probabilities in a single forward pass and speaks TypeSafe's System One API, so the TypeSafe SDK works unchanged. It self-hosts on roughly a 16 to 24 GB GPU or a 32 GB Apple Silicon Mac; all published results are author-reported.[1][2]
Strengths
- Answers typed questions (choice, score, yes/no) with a calibrated probability per option in a single forward pass, and the model card reports low calibration error across several held-out suites.[2]
- Implements TypeSafe's System One API as a drop-in, self-hostable endpoint that runs on a single GPU with about 16 GB free or a 32 GB Apple Silicon Mac.[2]
- Apache-2.0 open weights released as a LoRA adapter, intended by the author as a starting point for fine-tuning on a user's own labels.[2]
Limitations
- English only and not a generative model: it cannot chat, summarise, or answer open-ended questions, only score the options it is given; the author also excludes fully automated high-stakes decisions without human review.[2]
- Only 8,192 tokens of context are validated, the length the OpenRouter catalog also lists; states up to 65,536 tokens are served but reported accuracy falls at longer lengths, and the card flags knowledge-heavy tasks as weak.[1][2]
- The evidence here is the author's own model card plus a catalog listing, with no independent evaluation in this source pack; the card itself says confidence thresholds should be frozen on a labelled sample of the user's workload.[1][2]
Where to use it
- Document classification, routing, triage, extraction choices, and policy or eligibility checks over inputs of a few thousand tokens, per the author's stated intended uses and not independently verified.[2]
- Confidence-gated automation where likely cases are handled automatically and uncertain ones are queued for human review.[2]
- A self-hosted, API-compatible stand-in for a System One endpoint, or a base for fine-tuning decision tasks on proprietary labels.[2]
Sources
- [1] OpenRouter catalog · Jared Palmer: Kev 4BSources checked: Nov 13, 2026
- [2] Model documentation · Jared Palmer: Kev 4BSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Kev 4B is a small open-weight decision model from Jared Palmer, built as a LoRA adapter and pointer head on Qwen3.5-4B-Base and served over the same /v1/systemone contract as TypeSafe's...
Specifications
- Model ID
- jaredpalmer/kev-4b
- Canonical ID
- jaredpalmer/kev-4b-20260924
- Creator
- jaredpalmer
- Added to OpenRouter
- 25.09.2026
- Context length
- 8,192
- Modality
- text->decisions
- Input modalities
- text
- Output modalities
- decisions
- Token pricing
- No per-token price in the source snapshot — see OpenRouter for the billing unit
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | — |
| Coding | No measurement | — |
| Agents | No measurement | — |
| Writing | No measurement | — |
| Reasoning | No measurement | — |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
No evaluations are supplied for this model.
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| SiliconFlowfp8 | 8,192 | — | — | — | — | — |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.