Jev directory
CommunityJev Models & Reimplementations
Ollaya Models
Ollaya's searchable registry of open decision models you can run locally: 21 models with capability filters, and cards showing authors, sizes, tags, update dates, and typed-decision scores with RTX 4090 latency.
- Reading time
- 2 min
- Published by
- Community
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Capabilities
- Open sourceRuns locally
Highlights
- 21 models with capability filters including multilingual, router, guardrails, fine-tuned, decision, embedding, GGUF, long-context, and vision.
- Winnow by EldanRing fine-tunes Gemma 4 into GGUF models at 7.5b and 12b, scoring 0.722 typed decisions at 89 ms on an RTX 4090.
- Laya from Convai Innovations returns typed answers in one forward pass in English and 100+ languages, at 9.6 ms on an RTX 4090.
- KeV by Jared Palmer is a LoRA on Qwen3.5 with a pointer head that scores every option at its own span, at 0.669 and 354 ms.
- Also listed: decider by Mapika (0.680 at 4b), nli by Moritz Laurer (0.548 at 20.4 ms), and gliclass by Knowledgator (0.477).
Watch out
Grounded in the truncated model-search capture and the cobanov listing reviewed on 2026-10-15; the typed-decision scores and latencies are the registry's own measurements on an RTX 4090, several model cards beyond the first few were cut off, and no licences are shown.
Questions about this entry
Short answers drawn from the linked project or page.
Ollaya's model search: an index of open decision models you can run locally with the Ollaya runtime, with filters and a card per model showing authors, sizes, tags, scores, and latency measured on an RTX 4090.
21 at review time, with capability filters for multilingual, router, guardrails, fine-tuned, decision, embedding, GGUF, long-context, vision, and zero-shot, and sorting by popularity or newest.
Each card carries a typed-decisions score and a latency measured on an RTX 4090, for example winnow at 0.722 and 89 ms, or kev at 0.669 and 354 ms. They are the registry's own measurements, not third-party runs.
Among them: winnow (Gemma 4 fine-tunes by EldanRing at 7.5b and 12b), laya (Convai Innovations), decider (Mapika on Qwen3.5), kev (Jared Palmer), nli (Moritz Laurer), and gliclass (Knowledgator).
The registry calls them open decision models you can run locally with Ollaya. The capture does not show per-model licences, so check each model's own repository and licence before production use.
They are served through the Ollaya runtime, which downloads and serves decision models locally. The docs index covers pull, serve, and Modelfile workflows, though those pages were not part of the capture.
It is the dedicated model search rather than the marketing homepage: it adds filtering across 21 models plus per-model authors, sizes, update dates, and hardware-specific latency figures for each card.
More Jev resources like this
Open, CPU-sized Jev-style decision model: give it a situation, a question and options, and Decima returns calibrated probabilities in roughly 20-50 ms per decision as int8 ONNX, with no GPU, no API key and nothing leaving the machine.
Jev Models & Reimplementations#community#catalog-import#cobanov
Community
amyrmahdy/decimaPython
amyrmahdy/decimaPythonA 144.3M-parameter bilingual typed decision model fine-tuned from jhu-clsp/mmBERT-small: it answers noul, choice, and score questions in one forward pass with calibrated confidence and zero output tokens, and it runs on CPU.
Jev Models & Reimplementations#community#catalog-import#cobanov
An independent fork of Laya that replaces its text encoder with a small vision-language model, giving calibrated typed decisions over an image plus text in one forward pass from a 201M-parameter SmolVLM-256M backbone.
Jev Models & Reimplementations#community#catalog-import#cobanov
Community
r33drichards/laya-visionPython
r33drichards/laya-visionPythonA Japanese System One decision model built on modernbert-ja-310m: send text and typed questions and get typed answers with probabilities from one forward pass, no text generation, at 314.6M parameters that also run on a CPU.
Jev Models & Reimplementations#community#catalog-import#cobanov
Community
hiroki-abe-58/sokudanPython
hiroki-abe-58/sokudanPythonAlex Nikolic's open reproduction of Jev as a Qwen3.5 cross-encoder: one NLI primitive over premise and hypothesis - entailment, contradiction or neutral - that reranks answers, grades them, guards content and even picks game moves, with the 4B v5 checkpoint near Jev 1.13 on public JevBench items.
Jev Models & Reimplementations#community#catalog-import#madewithjev
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/ollayaRustFrom 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.