Jev directory
CommunityJev Models & Reimplementations
StartLux-Decision v1.1 9B Instant GGUF
StartLux's 9B Instant decision model as GGUF for llama.cpp: Q8_0 (9.53 GB) measures DecisionPivot 53.2 against 50.6 for Q4_K_M and 53.2 for BF16 on vLLM, returning one-forward-pass typed-question probabilities.
- Reading time
- 2 min
- Published by
- Community
- Author
- StartLux Models
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Shell
- Capabilities
- Self-hostedRuns locally
Highlights
- StartLux-Decision-v1.1-9B-Instant in Q8_0 (9.53 GB) and Q4_K_M (5.63 GB) for llama.cpp, plus a 918 MB F16 mmproj encoder.
- DecisionPivot on the 7,800 core items: Q8_0 53.2, Q4_K_M 50.6 and BF16-on-vLLM 53.2; Q8_0 is the recommended file.
- Instant models read prompts bidirectionally, so the files carry llama.cpp's non-causal attention flag.
- The card says default server batching cost the 4B Instant model 7.7 DecisionPivot points, and reading causally loses 27 of its 48 points.
- Text decoder only, without the BF16 checkpoint's multi-token-prediction layer.
Quickstart
bash
llama-server -m ~/models/StartLux-Decision-v1.1-9B-Instant-GGUF/StartLux-Decision-v1.1-9B-Instant-Q8_0.gguf -ngl 99 -c 20480 \
--parallel 1 --ctx-checkpoints 0 -b 16384 -ub 16384 --port 8081 &
STARTLUX_LLAMACPP_BIDIR=server startlux-decide serve --model ~/models/StartLux-Decision-v1.1-9B-Instant \
--backend llamacpp --llamacpp-url http://127.0.0.1:8081 --port 8090Watch out
Card reviewed 2026-10-17; DecisionPivot figures come from StartLux's own report (arXiv 2610.17309) with GGUF measurements on llama.cpp b11549, not third-party reproduction. Instant models read prompts bidirectionally, so default batching loses accuracy, and the licence is CC BY-NC 4.0.
Questions about this entry
Short answers drawn from the linked project or page.
StartLux-Decision-v1.1-9B-Instant in Q8_0 (9.53 GB) and Q4_K_M (5.63 GB) for llama.cpp, plus a 918 MB F16 vision encoder; a 9B decision model returning typed-question probabilities from one forward pass.
A state (text or JSON) plus typed questions, including pick-an-option, yes or no, and a level on a scale of 2 to 10, returns a probability for every option of every question from one forward pass in the TypeSafe /v1/systemone format.
DecisionPivot on 7,800 core items: Q8_0 53.2, Q4_K_M 50.6 and BF16 on vLLM 53.2; the card recommends Q8_0 and notes Q4_K_M lost up to 3.8 points across sizes.
Instant models read prompts bidirectionally in full-attention layers, so the files carry llama.cpp's non-causal attention flag and every prompt must be evaluated alone and in one piece; default batching cost the 4B Instant model 7.7 DecisionPivot points.
Run llama-server with -ngl 99 -c 20480 --parallel 1 --ctx-checkpoints 0 -b 16384 -ub 16384, set STARTLUX_LLAMACPP_BIDIR=server and start startlux-decide serve with the llama.cpp backend; tested with b11549.
Yes, through the separate mmproj F16 file (918 MB) passed to llama-server; the encoder is not quantized, so Q8_0 and Q4_K_M use the same file, while the GGUF itself holds the text decoder only.
CC BY-NC 4.0, non-commercial use only; the numbers come from StartLux's own report (arXiv 2610.17309) with GGUF measurements on llama.cpp b11549 rather than third-party reproduction.
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
TurinTech's independent GGUF conversion of the 27B Kev decision model for llama.cpp: the default IQ4_XS file is 13.91 GB and keeps 97.8% of the BF16 original's answers over 2,000 test decisions, with zero changed confident answers.
Jev Models & Reimplementations#community#open-models#local
Community
TurinTech's independent GGUF conversion of Jared Palmer's Kev-4B decision model; the default Q4_K_M file is 2.54 GB and returns the same answer as the BF16 original on 93.0% of 2,000 test decisions with no confident-answer changes.
Jev Models & Reimplementations#community#open-models#quantized
Community
TurinTech's independent GGUF conversion of Jared Palmer's Kev-9B decision model; the default IQ3_M file is 4.78 GB and matches the BF16 original on 92.2% of 2,000 test decisions with a minus 1.2-point accuracy change.
Jev Models & Reimplementations#community#open-models#quantized
Community
A Q8_0 GGUF of knowlet's Gemma-4-12B-Unified-System-One with a native System One decision runtime and an independently fitted GGUF temperature, shipped with an F16 mmproj for the model's native image and audio projections.
Jev Models & Reimplementations#community#open-models#quantized
Community
A 27B decision model from StartLux Labs that takes text, JSON or images up to 262,144 tokens and returns a probability per option through the TypeSafe /v1/systemone format, reporting Decision Index 0.2.1 63.88 and 208 of 231 on JevBench public.
Jev Models & Reimplementations#community#open-models#local
Community
The Q8_0 8-bit GGUF build of StartLux-Decision-v1-27B, 28.60 GB for llama.cpp, served through a decision wrapper that the card measures at 100% same decisions as the original and 209 of 231 on JevBench public.
Jev Models & Reimplementations#community#open-models#quantized
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.