Jev directory
CommunityJev Models & Reimplementations
MiniCPM5-2B-Jev on Ollama
ActBro's 2B calibrated System 1 decision model on Ollama, built on OpenBMB's MiniCPM5-2B: 78.79% on JevBench among sub-2B open-weight models, 86.41% on the 10-source decision-v7 set, and one output token per decision via /v1/systemone.
- Reading time
- 2 min
- Published by
- Community
- Author
- ActBro
- Last verified
- Oct 2026
- Platforms
- CLI
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- 2B single-pass calibrated System 1 decision model built on OpenBMB's MiniCPM5-2B, published under the ActBro namespace.
- Ranks #1 among all open-weight models of 2B parameters or fewer on JevBench, at 78.79% accuracy.
- 86.41% on the 10-source decision-v7 benchmark, above TypeSafe Hosted Jev at 84.50% and Together AI's Tev1 4B at 73.3%.
- Completes a decision in one output token, with sub-100 ms response times on consumer hardware, laptops, and Apple Silicon.
- 5.0 GB checkpoint with a 128K context window; native /v1/systemone plus a classic /api/generate compatibility mode and a standalone serve.py.
Watch out
The page is a community upload that calls the model open source and links a Hugging Face repo and GitHub project, but it names no licence for the weights, and the benchmark figures are the uploader's own claims without evaluation scripts or third-party runs. The page showed 103 downloads at capture, and the standalone path needs the extra calibration layer and PyTorch serving code.
Questions about this entry
Short answers drawn from the linked project or page.
A 2B single-pass calibrated System 1 decision model built on OpenBMB's MiniCPM5-2B and published on Ollama under the ActBro namespace. It is trained for support ticket triage, content moderation, policy enforcement, and agent decision routing.
Pull it with ollama pull actbro/minicpm5-2b-jev:2b and call Ollama 0.35 or later's /v1/systemone endpoint with a state and typed questions. The page also documents a classic /api/generate compatibility path for legacy pipelines.
The readme calls the model open source and links ytbai/MiniCPM5-2B-Jev on Hugging Face and yuting-ai/minicpm5-2b-jev on GitHub, but the Ollama page itself does not name a licence for the weights.
It ranks #1 among all open-weight models of 2B parameters or fewer with 78.79% accuracy, per the uploader's benchmark highlights. The page does not link a public JevBench leaderboard or evaluation script.
The page reports 86.41% accuracy on the 10-source decision-v7 benchmark, above proprietary TypeSafe Hosted Jev at 84.50% and Together AI's Tev1 4B at 73.3%, which is the uploader's headline comparison claim.
Single-pass: decisions complete in one output token, which the page says achieves sub-100 ms response times. The example response shows usage of 146 input tokens and 1 output token, with probabilities and confidence per answer.
Yes. The page documents standalone PyTorch serving: clone yuting-ai/minicpm5-2b-jev, install requirements, and run serve.py on a chosen port, aimed at distributed PyTorch clusters or raw CUDA/MPS setups with the post-hoc calibration layer.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
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 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
Bespoke Labs' 9B decision model on Ollama, fine-tuned from Qwen3.5-9B: Apache-2.0, up to 64 typed questions per call, answers in under 100 ms on a MacBook Pro with an M5 Max, and 75.7% mean accuracy across 13 human-labelled datasets.
Jev Models & Reimplementations#community#open-models#local
Community
Cloudflare's 9B multimodal decision model on Ollama: text, JSON, or images in one non-autoregressive forward pass, Apache-2.0, and the lowest measured latency of any decision model in Cloudflare's Decision Index runs at 38.8 ms median.
Jev Models & Reimplementations#community#open-models#local
Community
Convai Innovations' 421M non-autoregressive decision model on Ollama: fine-tuned from ModernBERT-large, Apache-2.0, answering one question in under 10 ms on an M5 Max with 0.060 in-task calibration error.
Jev Models & Reimplementations#community#open-models#local
Community
Together AI's experimental decision family on Ollama, fine-tuned from Qwen3.5 at 4B and 0.8B: 93K downloads, up to 64 questions per call, an 88.0% development accuracy on 1,000 decisions, and MIT-licensed dataset builders and training scripts.
Jev Models & Reimplementations#community#open-models#local
Community
Open Spark Jev's Qwen3.5-4B decision model on Ollama: one forward pass returns a calibrated probability per option for allow, ask, or deny tool gating, with 4b-v6 and 4b-v8 tags carrying fitted temperatures of 1.481 and 2.5 under Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
The typed-decisions checkpoint of the Laya family on ModelScope: a 421M ModernBERT-large fine-tune reporting 0.766 accuracy over 2,000 decisions, above TypeSafe Jev 1.13.0's published 0.727 and the 0.735 teacher ceiling, Apache-2.0.
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.