Up to date
jaredpalmer Apache 2.0 out 37 days ago
Text in, labels out.
Verified 2026-10-26 07:27 UTC Rank #0 in Decision models · score 0 method v1 Sources 3 Downloads 30d 20k · #458 of 619 tracked
Picked by rule from 1 tracked piece: the lab's own post, the most shared, a field report, the latest.
Kev is a family of small decision models built on Qwen3.5 and Qwen3.8 that you can train and run locally.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 12 quantized builds and 17 Spaces.
Top Spaces running it: multimodalart/jev-decision-index ♥ 802 · jaredpalmer/kev ♥ 93 · Station-house/ConspiracyBench ♥ 9 · jaredpalmer/kev-4b-decision-demo ♥ 2
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-09-19), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.
23 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
Kev-4B is an English decision model that functions as a LoRA adapter and pointer head on the Qwen3.5-4B-Base backbone.[1] The model is designed for text classification tasks such as routing and triage, returning calibrated probability distributions without generating text.[2] It is released under the Apache 2.0 license and can be served on a single 24 GB GPU.[3][4]
Generated 2026-10-07 15:18 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
Listed in Decision models
| Pipeline tag | text-classification |
|---|---|
| Library | peft |
| MoE | no |
| Languages | en |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-26 |
| How it was made | LoRA + new scoring/pointer head |
|---|---|
| Lineage bucket | F - fine-tune of an open model (public/synthetic data) |
| Lineage evidence | board file inspection (adapter/head/weight files) + HF base_model field |
| Base model | Qwen/Qwen3.5-4B-Base |
| DI 0.2.1 score | 34.64 (independent (DI 0.2.1 board)) |
| ECE (board) | 0.1756 |
| Median latency | 52.1 ms (DI board, 1x RTX PRO 6000, single process) |
| Hosted price (input) | self-hosted/M |
| Source | https://huggingface.co/jaredpalmer/kev-4b |
Base model: Qwen/Qwen3.5-4B-Base
quantized: 12adapter: 6finetune: 3merge: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| ggml-org/Kev-4B-GGUF | quantized | GGUF | — | — | apache-2.0 | ggml-org | 28 | 50.9k |
| mys/kev-4b-GGUF | quantized | GGUF | — | — | apache-2.0 | mys | 1 | 2.1k |
| turintech/Kev-4B-Artemis-GGUF | quantized | GGUF | — | — | apache-2.0 | turintech | 8 | 1.6k |
| RoderickQiu/kev-4b-mlx-8bit | quantized | mlx | 4.21B | M | apache-2.0 | RoderickQiu | 1 | 1.1k |
| Jakevin/kev-4b-ternary-mlx | quantized | mlx | 488M | S | apache-2.0 | Jakevin | 1 | 414 |
| alanhuangya/OpenJev-4B | adapter | — | 4.21B | M | apache-2.0 | alanhuangya | 2 | 232 |
| DreamBlooms/kev-4b-GGUF | quantized | GGUF | — | — | apache-2.0 | DreamBlooms | 1 | 252 |
| litert-community/Kev-4B-LiteRT | finetune | — | — | — | apache-2.0 | litert-community | 0 | 283 |
| aselea/Kev-4B-MLX-Serve-8bit | quantized | MLX | 4.21B | M | apache-2.0 | aselea | 0 | 245 |
| midudev/kev-4b-ONNX | quantized | — | — | — | apache-2.0 | midudev | 2 | 0 |
| ayan4m1/kev-4b-merged | merge | — | 2.66B | M | apache-2.0 | ayan4m1 | 0 | 87 |
| Flolep2607/kev-4b-actions | adapter | — | — | — | apache-2.0 | Flolep2607 | 1 | 0 |
| ArthurZ/kev-4b | finetune | — | — | — | apache-2.0 | ArthurZ | 1 | 0 |
| jtatman/kev-4b-code-verify-v1 | adapter | — | — | — | apache-2.0 | jtatman | 0 | 38 |
| Datchthana/kev-thai-pdpa-guard | adapter | — | — | — | apache-2.0 | Datchthana | 0 | 35 |
| jtatman/kev-4b-code-verify-v2 | adapter | — | — | — | apache-2.0 | jtatman | 0 | 31 |
| mlboydaisuke/Kev-4B-CoreAI | quantized | — | — | — | apache-2.0 | mlboydaisuke | 0 | 24 |
| danielamitay/kev-4b-fp32-swev | quantized | — | — | — | apache-2.0 | danielamitay | 0 | 7 |
| ai-ecoverse/kev-4b-vision-intent | quantized | — | — | — | openrail | ai-ecoverse | 0 | 0 |
| Flolep2607/kev-4b-guard | adapter | — | — | — | apache-2.0 | Flolep2607 | 0 | 0 |
| verathos-ai/kev-4b-mlx-fp16 | finetune | mlx | 4.21B | M | apache-2.0 | verathos-ai | 0 | 0 |
| Starlinghq/kev-4b-onnx | merge | — | — | — | apache-2.0 | Starlinghq | 0 | 0 |
| jfrog-boost/Kev-4B-GGUF | quantized | GGUF | — | — | apache-2.0 | jfrog-boost | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-10-01 | Kev 1.0 model card (weights unchanged) | 6cfce5c |
| 2026-09-24 | Kev-4B round 10: skills delta (hard-v1 + devtools-v1) on top of round 8 | 139fdd9 |
| 2026-09-24 | card: name the branch that holds the round-8 record | 1da696f |
| 2026-09-24 | Kev-4B: real-document delta (round 8; documents-v1 locked test 0.804 -> 0.904) | 957b91e |
| 2026-09-21 | card wording | 485ace8 |
| 2026-09-21 | calibration built in: head.pt carries the fitted temperature; KEV_TEMPERATURE=1.0 for raw logits | e226ccc |
| 2026-09-21 | Model card: link the Space demo | 70dd408 |
| 2026-09-21 | Kev-4B: dates+unknowable delta on the v7 checkpoint (locked test 0.871 / 0.837); previous weights at tag v7-base | 4bc64c6 |
| 2026-09-20 | Kev-4B: Qwen3.5-4B-Base, decision-v7 recipe (locked test 0.870 / 0.832); previous Qwen3 weights at tag qwen3 | f780095 |
| 2026-09-20 | Card: Kev display names; Kev-0.5B scored on the same out-of-domain items (0.561) | c4bfa11 |
| 2026-09-20 | kev-4b: card note | 4ccdc4e |
| 2026-09-20 | kev-4b: card refresh (0.6b column) | 6b332b5 |
| 2026-09-20 | kev kev-4b: release card | 55dc64b |
| 2026-09-20 | kev-4b preview, decision-v7 recipe (transfer 0.790 dev / 0.806 locked; held-out pairs 0.73) | 2bd3bb9 |
| 2026-09-19 | Card: Jev reference numbers from the transfer-v4 report (0.857 / Brier 0.211 / pairs 0.86) | 0d68f24 |
| 2026-09-19 | Card: third seed (0.759/0.758/0.759) | 002ed9f |
| 2026-09-19 | Card: fix per-source line | f9e2c84 |
| 2026-09-19 | Card: product-example caveat | d11d4fa |
| 2026-09-19 | Research preview: Qwen3-4B-Base + LoRA r16, decision-v4, lr 5e-5 (trial lowdrift-4b-v4/01-trial-1); dev 0.843, transfer 0.759; locked test 0.852 / 0.794 (one ungated read) | 1a0cb0a |
| 2026-09-19 | initial commit | e254cf5 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
Hub snapshot: 2026-10-26 · HF API: 2026-10-26 07:25 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:18 UTC · Last verified 2026-10-26 07:27 UTC
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
