Jev directory
CommunityJev Models & Reimplementations
Kev-27B
Jared Palmer's 27B Kev decision model: a full-weight fine-tune of Qwen3.8-27B averaged with v1, plus a pointer head that returns calibrated probabilities for typed choice, score and noul questions in one forward pass. Card-reported transfer-v4 accuracy is 88.87%.
- Reading time
- 3 min
- Published by
- Community
- Author
- Jared Palmer
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Platforms
- API
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- v2 is a full-weight fine-tune of Qwen/Qwen3.8-27B averaged with v1; the pointer head adds 2.6M parameters to a 25.6B backbone.
- One bf16 checkpoint at 51.3 GB serves states up to 65,536 tokens plus at least 8,192 tokens per question.
- Card-reported test accuracy: 0.8887 on transfer-v4 (656 questions) and 0.889 on hard, devtools and documents combined.
- Calibration is one stored temperature, T = 1.32, applied at load time; Apache-2.0 covers weights, head and base model.
- Implements TypeSafe's public POST /v1/systemone API, so the TypeSafe SDK runs against it unchanged.
Watch out
Card reviewed 2026-10-17; the accuracy and Brier figures are self-reported on the card, whose fetched text truncates the out-of-scope section after ruling out text generation and fully automated decisions with legal or similar consequences. Training states were at most 32,768 tokens, and serving hardware and latency are not stated.
Questions about this entry
Short answers drawn from the linked project or page.
A decision model from Jared Palmer that reads one document (the state) and typed questions and returns a calibrated probability distribution over the supplied options in one forward pass, without generating text. It targets document classification, routing, triage and policy checks.
The card says it implements TypeSafe's public System One API (POST /v1/systemone), so the TypeSafe SDK works against it unchanged. It is a self-hosted drop-in replacement for a System One endpoint, not affiliated with TypeSafe and not Jev itself.
Choice with 1 to 255 named options, score with 1 to 255 ordered levels, and noul yes/no questions. Each question is answered as its own row continuing from the shared cached state, so questions cannot influence one another.
The backbone holds 25.6B of the parameters (vision tower, LM head and multi-token-prediction layers are not loaded) plus a 2.6M-parameter head, in a 51.3 GB bf16 checkpoint. States up to 65,536 tokens are served, and the validated context length is 65,536 tokens.
Four held-out suites: transfer-v4 accuracy 0.8887 with Brier 0.154, breadth-v1 0.832, tasksource-heldout-v1 0.795 and hard plus devtools plus documents 0.889. All are self-reported on the card; no third-party rerun is linked.
Jared Palmer (github.com/jaredpalmer/kev). v2 is a full-weight fine-tune of Qwen/Qwen3.8-27B averaged with version 1, which was a rank-16 LoRA on the same base and remains at the v1-lora tag with temperature 1.38.
Text generation, chat, summarisation and open-ended question answering, since the model only scores the options it is given. The card's out-of-scope list also begins to rule out fully automated decisions with legal or similar consequences without human review, then truncates in the fetched text.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
Jared Palmer's 9B Kev decision model: a rank-16 LoRA and pointer head on a frozen Qwen3.5-9B-Base that answers typed choice, score and noul questions with calibrated probabilities in one forward pass, fitting one 24 GB-class GPU.
Jev Models & Reimplementations#community#open-models#local
Community
The smallest Kev: an 11.3M-parameter LoRA and pointer head on Qwen3.5-0.8B-Base that answers typed choice, score and noul questions with calibrated probabilities in one forward pass, running on a laptop or a 4 GB GPU.
Jev Models & Reimplementations#community#open-models#local
Community
Jared Palmer's Kev-4B: a 33.8M-parameter rank-16 LoRA and pointer head on Qwen3.5-4B-Base that returns calibrated probabilities for typed choice, score and noul questions in one forward pass, sized for one 24 GB GPU or a 32 GB Apple Silicon Mac.
Jev Models & Reimplementations#community#open-models#local
Community
Standard Thinking's 3B decision model with a joint schema head: it scores every option of every question in a scenario in one forward pass through POST /v1/systemone, and v3 self-scores 51.53 on the Decision Index 0.3 public index, up from v2's 42.90.
Jev Models & Reimplementations#community#open-models#local
Community
Standard Thinking's 8B decision model with a joint schema head that reads the scenario once and scores every option of every question; v4 self-scores 57.40 on the Decision Index 0.3 public index, up from v3's 48.34, with a 70.6M-parameter head.
Jev Models & Reimplementations#community#open-models#local
Community
The smallest decider from Mapika: a 1.4 GB bf16 Qwen3.5-0.8B-Base fine-tune that returns calibrated probabilities for typed Choice, Score and Noul questions in one forward pass and runs 1.5x faster than decider-2b, within one to four points on routing and lookups.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.