Jev directory
CommunityJev Models & Reimplementations
Standard One 8B LoRA
Standard Thinking's 8B LoRA adapter for the Standard One decision model: merge it with the supplied recipe and serve it, and it scores choice, noul and score labels over a scenario with calibrated probabilities via POST /v1/systemone, never generating text.
- Reading time
- 3 min
- Published by
- Community
- Author
- Standard Thinking
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- LoRA adapter plus merge recipe on mistralai/Ministral-3-8B-Instruct-2512-BF16; serving needs a merged checkpoint and the server code.
- Scores choice, noul and score labels in one forward pass and returns probabilities via POST /v1/systemone, with no generated text.
- v2.2 (2026-10-04) continues training from v2.1; earlier weights remain available under the tags v1, v1.1 and v2.
- On v2, mean accuracy was 76.6% with native wording and 76.4% with served wording over 10 held-out suites, none a JevBench tier.
- Apache-2.0 throughout, and the merged checkpoint retains the Pixtral vision tower; no image-input benchmark is reported.
Quickstart
python
import torch
from transformers import Mistral3ForConditionalGeneration
from peft import PeftModel
base = Mistral3ForConditionalGeneration.from_pretrained(
"mistralai/Ministral-3-8B-Instruct-2512-BF16",
revision="f6fae9795746f63c9be8344932f01275f3c63734",
torch_dtype=torch.bfloat16,
)
model = PeftModel.from_pretrained(base, ".").merge_and_unload()
model.save_pretrained("./StandardOne-8B-merged", safe_serialization=True)Watch out
Card reviewed 2026-10-17; the 76.6% and 76.4% wording figures are self-reported v2 measurements, the six-suite base comparison is offline rather than served-endpoint, and the historical v2 chart combines measurement paths. Serving requires merging the adapter and using StandardOne-8B's server code.
Questions about this entry
Short answers drawn from the linked project or page.
The 8B LoRA adapter of the Standard One decision model plus a merge recipe. It is not ready to serve on its own: you merge it into mistralai/Ministral-3-8B-Instruct-2512-BF16 and run the server code published with StandardOne-8B, which also hosts ready-to-serve merged BF16 weights.
Standard One scores a bounded set of answers for a supplied scenario and returns probabilities through POST /v1/systemone, without generating free-form text. That is the same typed-question, probability-answer shape Jev uses, though the card does not claim Jev compatibility beyond the endpoint.
Three types: choice selects among labeled options, noul is a yes/no decision, and score returns a distribution over an ordinal scale. The endpoint scores the supplied labels in one forward pass once the adapter is merged and served; answer text is never decoded.
v2.2 continues training from v2.1 with additional decision data, and questions with more than 26 options now use labels A to Z and then AA, AB. If you downloaded the adapter earlier, download it again or pin revision v2.2; v1, v1.1 and v2 stay available.
Merge the adapter with the card's Python recipe, then serve the merged directory with the sglang.launch_server and jev-adapter setup from StandardOne-8B's QUICKSTART.md, substituting the merged directory for --model-path, with model name standard-one-8b and no system prompt.
They are v2 measurements of mean accuracy over 10 suites, not JevBench tiers: native wording scored 76.6% and the served default 76.4%. No JevBench item was used to choose the wording or the temperatures, and the figures come from the card's own runs.
English plus Japanese, Chinese, Spanish, French, German, Portuguese and Russian, with a smaller Korean share. Apache-2.0 covers the base model, adapter, merged weights and server code; the merged checkpoint keeps the Pixtral vision tower and accepts image data URLs.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Both Jev 1.13 and 1.14 score 29 of 30 on 30 Turkish buyer and seller messages
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 NewsMore Jev resources like this
A LoRA plus 24-slot decision head over Qwen3.8-27B from Ming Xu's train-jev project: it serves TypeSafe's System One API with 0.851 Kev transfer-v4 accuracy and ECE 0.053, and needs an 80 GB GPU.
Jev Models & Reimplementations#community#open-models#local
Community
A 24 GB-GPU alternative from the train-jev project: Qwen3.5-4B plus a rank-16 LoRA and 24-slot decision head scoring 0.647 PK accuracy, 40.4 Decision Index and 43 ms p50 latency, ahead of Kev-4B at the same size.
Jev Models & Reimplementations#community#open-models#local
Community
Mapika's NVFP4 fine-tune of Gemma-4-31B-it answers typed Choice, Noul and Score questions with calibrated probabilities and no generated text, self-reporting a 64.27 chance-corrected public Decision Index and 39 ms median latency on a B300.
Jev Models & Reimplementations#community#open-models#local
Community
The v2.2 PEFT LoRA adapter for Standard Thinking's Standard One 3B decision model, shipped with a merge recipe for Ministral-3-3B-Instruct; serving requires the merged checkpoint and the shared server code in the StandardOne-8B repository.
Jev Models & Reimplementations#community#open-models#local
Community
Nox-4B, the 4B model in vLLM Semantic Router's Decision 1.0 family: give it a state, questions and possible answers, and it returns typed decisions and probabilities with labels defined at runtime. The card reports 73.09% weighted accuracy across 3,766 decisions.
Jev Models & Reimplementations#community#open-models#local
Community
The smallest decider from Mapika: a 1.4 GB bf16 Qwen3.5-0.8B-Base fine-tune that returns calibrated probabilities for typed Choice, Score and Noul questions in one forward pass and runs 1.5x faster than decider-2b, within one to four points on routing and lookups.
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.