Ecosystem
Quantization
Storing model weights at lower precision, 8-bit or 4-bit, to cut size and memory, at some risk to fine-grained probabilities.
- Category
- Ecosystem
- Also known as
- —
- Related terms
- 4
- Directory entries
- 3
- Docs
- —
- Added
- 2026-10-10
Definition
GGUF and ONNX builds make decision models practical on laptops and phones. Jev-Style v3's Q4_K_M file is 0.53 GB and, per its card, decision-identical to FP32 on 240 of 240 parity rows; Julia 1's ONNX export matched 100 of 100 WebGPU predictions with a maximum logit difference of 0.00225.
Rounding only flips near-ties in principle, but verify the artifact you deploy: bf16 serving in vLLM changed JEV-9B decisions in the third decimal depending on batching, which is why the AutoTrust cards apply head bias and temperature client-side.
Tagsopen-modelsmodels
Related terms
Definitions that connect to this one.
ONNXA portable model format that runs outside Python, including in browsers through WebGPU and in on-device runtimes.EcosystemOpen weightsModel checkpoints published for anyone to download, run, and fine-tune, as opposed to a hosted closed API like Jev.EcosystemBrier scoreA score for a full probability vector against one observed label. Lower is better, and a confident wrong answer costs more than a hedge.Evaluation & TrainingCalibrationThe property that outcomes given a probability of 0.2 occur about 20% of the time, measured across many predictions.Answers & Confidence
From the Jev directory
GGUF builds of the Jev-Style v3 decision model on Qwen3.5-0.8B: a 0.53 GB Q4_K_M file the card says is decision-identical to FP32 on 240/240 parity rows, served through a pinned llama.cpp scorer.
Jev Models & Reimplementations#community#open-models#local
Community
A browser WebGPU export of Julia 1: the same 551 MB decision checkpoint, a Rust WASM tokenizer, and a JavaScript API that answers typed questions client-side with no server or key.
Jev Models & Reimplementations#community#open-models#local
Community
The 9B member of AutoTrust's Jev students: a 40.2M-parameter decision block on Qwen3.5-9B with 0.019 mean KL to Jev 1.13 and 70.7% HumanEval left untouched, small enough for one modest GPU.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev community
Posts from builders shipping with Jev right now.