Jev directory
CommunityJev Models & Reimplementations
StartLux-Decision v1.1 0.8B Instant GGUF
GGUF quantizations of StartLux Labs' 0.8B Instant decision model for llama.cpp: Q8_0 (0.81 GB, DecisionPivot 9.7) and Q4_K_M (0.53 GB, 8.7), plus a 205 MB F16 vision encoder, against 9.5 for BF16 on vLLM.
- Reading time
- 2 min
- Published by
- Community
- Author
- StartLux Models
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Platforms
- API
- Capabilities
- Self-hostedRuns locally
Highlights
- Q8_0 file is 0.81 GB and scores 9.7 DecisionPivot on llama.cpp; Q4_K_M is 0.53 GB and scores 8.7; BF16 on vLLM scores 9.5.
- Vision encoder mmproj file is 205 MB, and the text-decoder files omit the multi-token-prediction layer of the BF16 checkpoint.
- Same bidirectional prompt requirement as its siblings: one slot, no context checkpoints and -b 16384 -ub 16384 on a stock llama-server.
- Weights are CC BY-NC 4.0; Q4_K_M lost up to 3.8 DecisionPivot points across the sizes; Q8_0 is the recommended file; tested with llama.cpp b11549.
Quickstart
bash
curl -s localhost:8090/v1/systemone -H 'Content-Type: application/json' -d '{
"state": {"ticket": "I was charged twice for order #4411 and the app still shows it as unpaid."},
"questions": {
"team": {"type": "choice", "instructions": "Which team should handle this ticket?",
"criteria": {"billing": "Payments and refunds", "shipping": "Delivery", "technical": "App and login"}},
"urgent": {"type": "noul", "instructions": "Should this ticket be answered today?"}
}
}'Watch out
Card reviewed 2026-10-17; the DecisionPivot scores are the publisher's own llama.cpp b11549 measurements, not third-party runs, and the gaps at this size are small: Q8_0 9.7 against the BF16 checkpoint's 9.5 on vLLM. Weights are CC BY-NC 4.0, so commercial use requires a separate licence, and plain chat with a GGUF file does not give these decisions without the package's prompt handling.
Questions about this entry
Short answers drawn from the linked project or page.
A llama.cpp GGUF package for StartLux-Decision-v1.1-0.8B-Instant, a 0.8B decision model from StartLux Labs that returns a probability for every option of every typed question from one forward pass without generating text.
Requests and responses use the TypeSafe /v1/systemone format, so it takes the same shape as Jev: a state plus typed questions, answered with probabilities over your options rather than generated text.
Q8_0 (0.81 GB) and Q4_K_M (0.53 GB) text-decoder files, plus the mmproj vision encoder at 205 MB. Q8_0 is the recommended file; Q4_K_M lost up to 3.8 DecisionPivot points across the family's sizes.
Download the GGUF plus the tokenizer and configs from the BF16 repository, install the StartLux-Decision package, and run llama-server with the bidirectional prompt options (one slot, no context checkpoints, -b 16384 -ub 16384) behind the startlux-decide wrapper. Tested with llama.cpp b11549.
Mean chance-corrected accuracy over the scorer's seven layers on 7,800 core items: 9.7 for Q8_0 on llama.cpp, 8.7 for Q4_K_M and 9.5 for the BF16 weights on vLLM. The numbers are the publisher's own measurements in its arXiv report.
StartLux Labs built it, with code at StartLuxLabs/StartLux-Decision and a paper at arXiv 2610.17309. The weights ship under CC BY-NC 4.0, so non-commercial use is free and commercial use needs a separate licence.
It is the smaller 0.8B sibling in the same GGUF series: 0.81 GB Q8_0 against 2.01 GB, with DecisionPivot 9.7 against 23.7, so it trades a large amount of accuracy for roughly a third of the disk and memory footprint.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
X: Standard One SH (StandardThinking)
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
GGUF quantizations of StartLux Labs' 2B Instant decision model for llama.cpp: Q8_0 (2.01 GB) and Q4_K_M (1.27 GB), plus a 668 MB F16 vision encoder, with Q8_0 scoring 23.7 DecisionPivot against 23.5 for BF16 on vLLM.
Jev Models & Reimplementations#community#open-models#local
Community
prithivMLmods' GGUF builds of autotrust/JEV-35B, the Qwen3.5-35B-A3B MoE decision model that returns calibrated choice, noul and score answers over text and images; files run from BF16 to Q3_K_L for llama.cpp.
Jev Models & Reimplementations#community#open-models#local
Community
A 4-bit W4A8 AWQ plus GPTQ build of AutoTrust's JEV-27B-VL from AI Outfitters that keeps the rank-32 LoRA decision head and a bf16 vision tower, served by vLLM on a single 24 GB GPU.
Jev Models & Reimplementations#community#open-models#local
Community
StartLux Labs' 27B AutoThink decision model reads a state (text, JSON or up to 8 images) and typed questions, answers in one pass and only writes a thought when the top probability falls below the 0.6 gate; self-reported Decision Index 0.3 public score 73.37.
Jev Models & Reimplementations#community#open-models#local
Community
StartLux Labs' 4B Instant decision model returns a probability for every option of every typed question from one forward pass with no generated text; it self-reports Decision Index 0.3 public 67.88 and DecisionPivot pair accuracy 41.5.
Jev Models & Reimplementations#community#open-models#local
Community
Mapika's NVFP4 fine-tune of Gemma-4-31B-it answers typed Choice, Noul and Score questions with calibrated probabilities and no generated text, self-reporting a 64.27 chance-corrected public Decision Index and 39 ms median latency on a B300.
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.