Jev directory
CommunityJev Models & Reimplementations
Standard One 8B
Standard Thinking's merged BF16 8B checkpoint of Standard One, in the repository that also carries the shared server code: bounded choice, noul and score questions get probabilities through POST /v1/systemone, with no generated answer text.
- Reading time
- 2 min
- Published by
- Community
- Author
- Standard Thinking
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- Merged BF16 8B checkpoint (base Ministral-3-8B-Instruct-2512-BF16) plus the shared jev-adapter server code.
- choice, noul and score questions return probabilities over the caller's labels in one forward pass, without decoding text.
- v2 card offline comparison: 8B beat its untuned base on four of six suites but declined on public easy (2.08) and public hard (4.50).
- Probabilities are temperature-scaled; calibration was checked on v2 (hard-tier ECE and distribution total variation).
- Training mixture spans nine languages; the Pixtral vision tower accepts image data URLs, with no image-input decision benchmark.
Quickstart
bash
.venv-native/bin/jev-adapter --engine-url http://127.0.0.1:30000 --model standard-one-8b --alias jev-latest \
--host 0.0.0.0 --port 30120 --max-concurrency 1 \
--tokenizer-model mistralai/Ministral-3-8B-Instruct-2512-BF16 \
--tokenizer-revision f6fae9795746f63c9be8344932f01275f3c63734 \
--prompt-wording served --label-scheme upper --default-temperature 1.65Watch out
Card reviewed 2026-10-17; the offline base-model comparison (four of six suites, declines on public easy and public hard) is explicitly not a served-endpoint result, and the calibration checks and prompt-wording accuracy table were measured on v2. No separate image-input decision benchmark is reported.
Questions about this entry
Short answers drawn from the linked project or page.
The merged BF16 8B checkpoint of Standard One, built on Ministral-3-8B-Instruct-2512-BF16, in the repository that also carries the shared server code. It returns probabilities for bounded choice, noul and score questions through POST /v1/systemone and never generates free-form text.
In the reported served evaluations, 8B scores higher than 3B on the public standard and hard tiers, while 3B has a lower median latency on the measured short-request profile. Both offer the same endpoint shape and label scheme.
Measured on v2, the 8B checkpoint improves on four of six suites against its untuned base but declines on public easy by 2.08 points and public hard by 4.50. The card stresses these are not served-endpoint results.
On a CUDA-capable Linux host, run stock SGLang 0.5.20 for the engine and jev-adapter in separate terminals, with --prompt-wording served --label-scheme upper --default-temperature 1.65 and the tokenizer revision f6fae9795746f63c9be8344932f01275f3c63734.
The training mixture covers English, Japanese, Chinese, Spanish, French, German, Portuguese, Russian and a smaller Korean share, and the retained Pixtral vision tower accepts image data URLs. No separate image-input decision benchmark is reported.
FP8 compressed-tensors weights validated with SGLang, and GGUF files for llama.cpp from BF16 down to IQ2_M with a vision projector. Validation numbers live in each repository's card.
Apache-2.0 covers the base model, adapter, merged weights and the shared server code hosted in the 8B repository; calibration was checked on v2 through hard-tier ECE and distribution total variation.
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
Standard Thinking's merged BF16 3B checkpoint of Standard One: it scores choice, noul and score questions and returns probabilities through POST /v1/systemone with no generated text. The v2.2 card reports 100.00% JevBench public easy and 79.97% on many-option questions.
Jev Models & Reimplementations#community#open-models#multilingual
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
FRIDA-Decisions answers Russian text with structured decisions in a single encoder pass — choice, ordinal score, yes/no or ranking — and scores 0.893 on the 735-item razvilka benchmark, effectively level with TypeSafe Jev's 0.897 (McNemar p = 0.84).
Jev Models & Reimplementations#community#open-models#russian
Community
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
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.
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
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.