CommunityJev Models & Reimplementations
JEV-9B GGUF
GGUF quantization of AutoTrust's JEV-9B Blocks-of-Experts model: eight files from BF16 17.9 GB down to Q3_K_M 4.62 GB, wrapping the frozen Qwen3.5-9B backbone and its System 1 decision head for llama.cpp.
- Published by
- Community
- Author
- prithivMLmods
- Last verified
- Oct 2026
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- GGUF quantization of AutoTrust's JEV-9B, a Blocks-of-Experts model on a frozen, bit-identical Qwen3.5-9B backbone.
- Eight files from BF16 at 17.9 GB down to Q3_K_M at 4.62 GB; the card marks Q4_K_M (5.63 GB) the default pick.
- System 2 is untouched base lm_head text generation at 70.7% HumanEval pass@1, with all 164 completions byte-identical to base.
- System 1 is a 40.2M-parameter LoRA plus a 24-slot fp32 decision head answering noul, choice, and score in one forward pass.
- On 25,376 held-out Jev-labelled rows: KL about 0.019 nats, 90.2% choice top-1, 0.994 noul AUROC, 0.0007 ECE.
Watch out
Apache-2.0 per the card, which states JEV-9B is an independent reproduction sharing no code or weights with TypeSafe's closed hosted Jev 1.13 and distills via the Apache-2.0 SargeDev/jev-distill-corpus-v3 corpus. The benchmark figures belong to AutoTrust's reference model, distill, and eval reports; this repo supplies the eight GGUF files but no quant-specific scores.
Questions about this entry
Short answers drawn from the linked project or page.
prithivMLmods' GGUF repackaging of AutoTrust's JEV-9B, an integrated System 1 and System 2 model whose frozen Qwen3.5-9B backbone serves ordinary text generation while a detachable System 1 head answers typed decisions.
Sizes run from BF16 at 17.9 GB through Q6_K 7.36 GB and Q5_K_M 6.47 GB to Q3_K_M at 4.62 GB. The card marks Q4_K_M (5.63 GB) the default size for most use cases and Q6_K near perfect.
System 2 is untouched base lm_head text generation at 70.7% HumanEval pass@1; System 1 is a 40.2M-parameter LoRA plus a 24-slot fp32 decision head that answers noul (yes/no), choice (2-16 options), and score (0-5) questions in one forward pass.
On 25,376 Jev-labelled held-out rows the card reports a mean KL divergence of about 0.019 nats from the teacher's distributions, 90.2% choice top-1 agreement, 0.994 noul AUROC, and 0.0007 ECE with no post-hoc correction.
The card reports about 90ms median per decision versus 238-301ms for the hosted service, and about 15x the throughput on one B200, with both systems served from one set of weights via a single vLLM engine and a per-call route.
System 2 generation, and the card says it is identical to base Qwen3.5-9B with all 164 completions byte-identical. The larger sibling autotrust/JEV-27B reaches 78.0% HumanEval but trades some speed and teacher fidelity for it.
No. AutoTrust AI states it is an independent, unaffiliated reproduction sharing no code or weights with TypeSafe's closed hosted Jev 1.13, distilled from that service's output distributions via the Apache-2.0 SargeDev/jev-distill-corpus-v3 corpus.
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
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 NewsHacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsMore Jev resources like this
Vision-capable GGUF port of autotrust/JEV-27B-VL: BF16 through Q3_K_M with a separate mmproj file, running the calibrated System 1 typed-decision interface and a 27B System 2 through llama.cpp.
Jev Models & Reimplementations#community#open-models#local
Community
chenz53's standalone decision model derived from Google Gemma 4 26B A4B IT: full merged BF16 weights, a fitted temperature of 1.736507, and 79.50% equal-type accuracy over 1,100 final questions.
Jev Models & Reimplementations#community#open-models#local
Community
1.3kGitHub stars
A local daemon that pulls open decision models and serves them on TypeSafe's /v1/systemone, so a Jev client switches hosts with TYPESAFE_BASE_URL.
Jev Models & Reimplementations#community#rust#open-models
Community
ollaya-dev/ollayaRust
ollaya-dev/ollayaRustThe NVFP4 build of AutoTrust's JEV-35B: routed experts quantized to 4-bit cut the download to 25.6 GB and GPU weights to 23.3 GiB, keeping 96.6% of bf16's Decision Index answers.
Jev Models & Reimplementations#community#open-models#local
Community
AutoTrust's combined System 1 and System 2 bundle on Gemma-4-26B-A4B-it: Decision Index 0.2.1 of 58.05, a 24-slot head with noul, choice and score, and the base chat model left bit-identical.
Jev Models & Reimplementations#community#open-models#local
Community
Reza2kn's llama.cpp Q4_K_M build of Jev-Omni: a 6.87 GiB multimodal decision backbone plus projector and FP32 head, measured at 8.98 GiB VRAM on an RTX 5080 Laptop GPU.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.
Jev Models & Reimplementations#open-models#training#local
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
Read article
From the Jev community
Posts from builders shipping with Jev right now.