Jev directory
CommunityJev Models & Reimplementations
Jev-Omni on Replicate
barakplasma's Replicate deployment of Jev-Omni: ask a typed yes/no or multiple-choice question about an image and get a calibrated probability per option from one forward pass, $0.00098 per run on L40S, Apache-2.0, and #1 on Image JevBench v0.1.
- Reading time
- 2 min
- Published by
- Community
- Author
- barakplasma
- Last verified
- Oct 2026
- Platforms
- API
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Pinned deployment of akhilaaa3/Jev-Omni, Apache-2.0, on Replicate under barakplasma; the listing claims #1 on Image JevBench v0.1.
- Ask a typed yes/no or multiple-choice question about an image and get a calibrated probability per option from one forward pass.
- Costs about $0.00098 per run, or 1,020 runs per dollar, on Nvidia L40S hardware; predictions typically complete within 1 second.
- Open source, per the page, which says you can also run it on your own computer with Docker.
- 1.4K public runs and no readme; the listing shows the model created 2 weeks and 4 days before capture and updated 2 weeks and 2 days before.
Watch out
Apache-2.0 per the listing, but the page carries no readme: inputs, image formats, question limits, and output schema are undocumented, and the Image JevBench #1 claim is not backed by a benchmark table. Cost and latency are Replicate's own estimates that vary with inputs, and the deployment pins a third-party revision of Jev-Omni rather than publishing its own weights.
Questions about this entry
Short answers drawn from the linked project or page.
A public Replicate deployment by barakplasma of Jev-Omni, pinned from akhilaaa3/Jev-Omni and listed as Apache-2.0. It takes a typed yes/no or multiple-choice question about an image and returns a calibrated probability per option from one forward pass.
Replicate estimates about $0.00098 per run, or 1,020 runs per dollar, varying with inputs. Runs execute on Nvidia L40S hardware and predictions typically complete within 1 second, according to the listing.
The listing states Jev-Omni was #1 on Image JevBench v0.1. No benchmark table, evaluation script, or methodology accompanies that claim on the page itself.
Yes, the page says the model is open source and can be run on your own computer with Docker, in addition to the hosted Replicate API. The pinned revision comes from the akhilaaa3/Jev-Omni account.
It shows 1.4K public runs. The listing records the model as created 2 weeks and 4 days before capture and updated 2 weeks and 2 days before, so it is a recent deployment.
There are none on this page: the listing has no readme, so inputs, image formats, question limits, and output schema are documented only by the one-line description, and integration work would need the pinned model's own documentation.
barakplasma publishes it on Replicate, while the underlying Jev-Omni revision is pinned from the akhilaaa3 account. The page shows no support channel, licence text beyond the Apache-2.0 label, or contact information.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker 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 NewsX: Happy JevDay: images, box, longer context, and an opt-in to find jaggedness
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
More 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
Singularity Compute's fine-tune of Google's diffusiongemma-26B-A4B-it, a discrete-diffusion mixture-of-experts, served from an 18 GB NVFP4 checkpoint that answers choice, noul and score questions with calibrated distributions on one 48 GB GPU.
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
AutoTrust's JEV-9B on ModelScope: a Blocks of Experts System 1 + System 2 model fine-tuned from Qwen3.5-9B, Apache-2.0, about 8.95 GB of BF16 safetensors in five shards, taking images since 3 October 2026, with a GuruSearch live demo.
Jev Models & Reimplementations#community#open-models#local
Community
AutoTrust's bf16 System 1 decision model on Qwen3.5-35B-A3B with about 3B active parameters: Decision Index 0.3 public 59.73, ERQA 51.3% in one pass with zero robot training, 66.5 GiB of weights.
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.