CommunityJev Models & Reimplementations
StartLux-Decision v1 4B
A 4B decision model taking text, JSON, or images up to 262,144 tokens: 52.75 on Decision Index 0.2.1, 204 of 231 on JevBench public, and 26.0 ms for a three-question request on one H200.
- Published by
- Community
- Author
- StartLux Labs
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Decision Index 0.2.1 scores 52.75 and Decision Index 0.2 scores 48.38, with 204 of 231 correct on JevBench public.
- 26.0 ms end to end for one request with three questions and 14.7 ms for one yes/no question, on one H200 in bf16.
- Native context of 262,144 tokens; a long state is read once in chunks and every question branches off it.
- Ships its own startlux_decision package; without flash-linear-attention and causal-conv1d the fallback is more than ten times slower.
- Requests and responses use the TypeSafe /v1/systemone format, so clients written for Jev work unchanged.
Watch out
The weights are CC BY-NC 4.0: free for research and non-commercial use, with commercial use requiring a separate licence from StartLux Labs, though the inference code in startlux_decision/ is Apache-2.0. The comparison table's figures for other systems are the public board's values, and the MLX backend and GGUF files read text only.
Questions about this entry
Short answers drawn from the linked project or page.
CC BY-NC 4.0: free for research and other non-commercial use with attribution, while commercial use requires a separate licence from StartLux Labs. The inference code in startlux_decision/ is Apache-2.0.
On one H200 in bf16, one request at a time: 26.0 ms end to end for a request with three questions and 14.7 ms for a single yes/no question, with CUDA graphs replayed for short requests.
Decision Index 0.2.1 scores 52.75 and Decision Index 0.2 scores 48.38; it answers 204 of 231 JevBench public items and averages 91.17 across Intern-Decision's seven suites.
Yes. The weights include a vision tower and the inference package uses it: pass images=[...] in Python or images over HTTP (base64 strings or data URIs), and mark placement with <image> in a string state. The MLX backend and GGUF files read text only.
hf download the repo, pip install -r requirements.txt, run python -m startlux_decision.check . and confirm it prints 'fast kernels: active', then python -m startlux_decision.server --model . --port 8090.
The card says transformers falls back to a path more than ten times slower, and the server refuses to start on a GPU without flash-linear-attention and causal-conv1d installed.
On JevBench public it answers 204 of 231 against Jev 1.13's 199 and JevK5's 200. Its Decision Index 0.2.1 of 52.75 sits below Jev 1.13's 57.91; the card notes the DI values come from the public board.
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
An unofficial cross-encoder fine-tune of ModernBERT-ja-310m for Jev Local: it reads a question-plus-state and one candidate per sequence and hits 92.62% on the JGLUE JNLI test, with CC BY-SA 4.0 weights.
Jev Models & Reimplementations#community#open-models#local
Community
akhilaaa3's merged multimodal decision classifier on Gemma 4 12B IT: probabilities for text, image, audio or video options, 87.57% on DecisionBench Medium and MMAU 63.10%, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Mithril's open, Jev-shaped typed-decision model on DeBERTa-v3-large: choice up to 255 options, score and noul from one forward pass, 0.854 in-domain with ECE 0.022, trained on public gold labels only.
Jev Models & Reimplementations#community#open-models#local
Community
The 322M mmBERT-base build of Laya for 100+ languages: state plus typed questions return typed answers with probabilities in one forward pass, with up to 8,192 tokens for long documents.
Jev Models & Reimplementations#community#open-models#local
Community
An Open Classification Foundation smoke checkpoint: a 164M-parameter ModernBERT-base with a cross-entropy decision head, 0.751 accuracy and 0.044 ECE on 2,000 held-out typed decisions.
Jev Models & Reimplementations#community#open-models#local
Community
An open Thai + English System One decision model at 0.8B, Apache-2.0: choice with up to 255 options, fractional score, and noul answers in one forward pass over a Qwen3.5-0.8B text tower.
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
Jev takes text only; Clef-omni adds image, audio, and video input and Liquid d1 adds images, each keeping the same typed-answer shape. Where that matters, where it does not, and the costs the vendors underplay.
Jev Models & Reimplementations#multimodal#open-models#decision-model
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
From the Jev community
Posts from builders shipping with Jev right now.