Jev directory
CommunityJev Models & Reimplementations
StartLux-Decision v1 27B
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.
- Reading time
- 2 min
- Published by
- Community
- Author
- StartLux Labs
- Last verified
- Oct 2026
- Languages
- Shell
- Capabilities
- Self-hostedRuns locally
Highlights
- CC BY-NC 4.0 weights; the family spans 0.8B, 2B, 4B, 9B, 27B and 35B-A3B on the same typed-decision interface.
- Card reports Decision Index 0.2.1 63.88, JevBench public 208 of 231 and 91.82 average over Intern-Decision's seven suites.
- One H200 in bf16: 102.3 ms end to end for three questions in one forward pass, 50.7 ms for a single yes/no.
- Ships its own startlux_decision package; without flash-linear-attention and causal-conv1d the server refuses to start on a GPU.
- Requests and responses use the TypeSafe /v1/systemone format, so clients written for Jev work unchanged.
Quickstart
bash
hf download startlux-models/StartLux-Decision-v1-27B --local-dir StartLux-Decision-v1-27B
cd StartLux-Decision-v1-27B
pip install -r requirements.txt
python -m startlux_decision.check .
python -m startlux_decision.server --model . --port 8090Watch out
Card reviewed 2026-10-17; the Decision Index, JevBench, Intern-Decision average and latency figures are all self-reported on the card, and the comparison rows for other systems are the public board's values. The weights are CC BY-NC 4.0, so commercial use requires a separate licence.
Questions about this entry
Short answers drawn from the linked project or page.
The 27B model of StartLux Labs' StartLux-Decision v1 family. It answers typed questions about a state and returns a probability for each option through the TypeSafe /v1/systemone format, so clients written for Jev work unchanged. Sibling sizes run from 0.8B to 35B-A3B.
Self-reported on the card: Decision Index 0.2.1 63.88 and 0.2 59.54, 208 correct answers of 231 JevBench public items, and 91.82 average accuracy over Intern-Decision's seven suites.
In bf16 with one request at a time: 102.3 ms end to end for a request carrying three questions, and 50.7 ms for a single yes/no question. All questions of a request share one forward pass, and the server replays recorded CUDA graphs for short requests.
Yes. The weights include a vision tower: pass images=[...] in Python or base64 and data URI images over HTTP, with <image> in a string state marking where each image goes. The MLX backend and the GGUF files read text only.
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 repository ships its own inference package, startlux_decision, as the fast path.
It is a competing decision model speaking the TypeSafe /v1/systemone contract. The card's table puts Jev 1.13 at 199 of 231 JevBench public and Decision Index 0.2.1 57.91, versus 208 and 63.88 here; those competitor values come from the public board.
Download the repository, pip install -r requirements.txt, run python -m startlux_decision.check . to confirm the fast kernels are active, then python -m startlux_decision.server --model . --port 8090 and POST to /v1/systemone.
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
The Ollama build of iapp's OpenThai-SystemOne: a 0.8B Thai and English decision model whose v0.3 weights are fine-tuned for Ollama's /v1/systemone letter readout, reporting a 74.3 macro across 13 public subsets for Q8_0.
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
The Q8_0 8-bit GGUF build of StartLux-Decision-v1-9B, 9.53 GB for llama.cpp, with the card reporting 100.0% same decisions as the original weights and 200 of 231 correct (86.6%) on JevBench public.
Jev Models & Reimplementations#community#open-models#quantized
Community
A 27B decision model from torchcast-ai that answers choice, noul and score questions together in one forward pass over the TypeSafe-compatible /v1/systemone format; the card reports Decision Index 0.2.1 of 65.00, +7.09 over Jev 1.13, at 33 ms per yes/no question.
Jev Models & Reimplementations#community#open-models#local
Community
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/ollayaRustA small distilled sibling of Perplexity's pplx-decider-v1.2: a Qwen3.5-4B decider that reads a state plus typed choice, noul and score questions in one pass, keeping most of the 27B model's Decision Index.
Jev Models & Reimplementations#community#open-models#multimodal
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.