Jev directory
CommunityJev Models & Reimplementations
Kev-27B Artemis GGUF
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.
- Reading time
- 3 min
- Published by
- Community
- Author
- TurinTech
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Shell
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- IQ4_XS default: 13.91 GB, 97.8% same answers as the BF16 original on 2,000 test decisions and zero changed confident answers.
- Q4_K_S (14.56 GB) tracks the original probabilities closest; Q3_K_M is smallest at 12.31 GB with 17.5 GiB peak memory.
- Runs on stock llama.cpp via the /v1/systemone endpoint; keep -ub 4096 -b 4096 or inputs longer than 512 tokens are rejected.
- Default file needs 19.0 GiB peak GPU memory; tested on an Apple M3 Max (36 GB) with Metal, where every file loads and answers.
- Independent TurinTech conversion of jaredpalmer/kev-27b under Apache-2.0, not affiliated with or endorsed by the Kev authors.
Quickstart
bash
llama-server -hf turintech/Kev-27B-Artemis-GGUF:IQ4_XS --port 8080 -c 16384 -ub 4096 -b 4096
curl http://localhost:8080/v1/systemone -H 'Content-Type: application/json' -d '{
"state": "My Stripe integration keeps failing. Please help ASAP.",
"questions": {"urgency": {"type": "noul", "instructions": "Does this message express urgency?"}}
}'Watch out
Card reviewed 2026-10-17; the same-answer rates, accuracy deltas, KL comparisons and the 19.0 GiB memory figure are the converter's own measurements on a 2,000-decision sample, and the card warns that the comparisons informed file selection and do not establish performance on unseen inputs.
Questions about this entry
Short answers drawn from the linked project or page.
TurinTech's conversion of the 27B jaredpalmer/kev-27b decision model to standard GGUF for llama.cpp. It accepts text with typed questions and returns a probability for every allowed answer in a single pass instead of generating text; it is text only.
The card recommends IQ4_XS (13.91 GB): 97.8% same answers as BF16 with zero changed confident answers. Q4_K_S (14.56 GB) keeps probabilities closest (KL .0029), Q3_K_M (12.31 GB) trades 0.15 accuracy points for 17.5 GiB memory, and the BF16 original is 51.3 GB.
Serve the GGUF with llama-server and POST state plus typed questions to /v1/systemone; the model answers each question with probabilities per allowed option. The card stresses this is a decision model, so use /v1/systemone rather than a chat prompt.
Run llama-server -hf turintech/Kev-27B-Artemis-GGUF:IQ4_XS --port 8080 -c 16384 -ub 4096 -b 4096, or point -m at a downloaded file. Keep -ub 4096 -b 4096 because the default of 512 rejects longer inputs. Tested with llama.cpp commit fcac190 from 13 October 2026.
The default file measures 19.0 GiB peak GPU memory, and the card tested every file on an Apple M3 Max (36 GB) with llama.cpp Metal. For comparison, the public Q6_K file is closer to the original but needs 25.7 GiB.
Yes: the card shows piping a message to ollama run turintech/kev-27b-artemis:13.9gb --ask with a decision question, tested on Ollama 0.40.4. The size tag selects the IQ4_XS file and is not an immutable revision; include the model name in the decision request to port 11434.
No. The card says it is an independent TurinTech conversion, not affiliated with or endorsed by the Kev authors; the files are standard GGUF and run on stock llama.cpp, Apache-2.0, with the weights converted and quantized for llama.cpp.
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
A community Q8_0 GGUF of Mapika/decider-0.8b, the smallest System One typed-decision model: one forward pass returns calibrated Choice, Score and noul answers, reporting 0.405 on JevBench hard-public at 920 MB resident.
Jev Models & Reimplementations#community#open-models#local
Community
An 8-bit MLX conversion by SirSahOl of chaoliangUNSW/Jev-Style-0.8B-Decision-v3 built for Apple Silicon native GPU inference: about 990 MB of active VRAM, a 262,144-token context and an average of 8.25 bits per weight.
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
Fluid Inference's fixed-shape Core ML conversion of the classification path of Fastino's GLiNER2.5-Decide for Apple silicon: the DeBERTa-v3-large encoder and shared label classifier score up to four decision heads per call, in fp16, 8-bit or 6-bit packages.
Jev Models & Reimplementations#community#open-models#local
Community
A CPU-first decision model from Decision Models that answers the seven questions an LLM auto-router needs (category, difficulty, stakes, four flags) in one ONNX pass; MIT weights plus a Jev-compatible hosted API.
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.