3 models · checked 20 September 2026

Decision models in ONNX

ONNX exports, for serving stacks that will not load PyTorch. Also what you want for browser and edge runtimes.

Also in this dimensionGGUF 7MLX 5Core ML 6
Community card

rlcd-modernbert-151m

heman10x

The smallest decision model we track, at 151M parameters on a GLiClass/ModernBERT encoder. If it holds accuracy, it changes what this class of model costs to run.

151M · knowledgator/gliclass-modern-base-v2.0 · apache-2.0

Community card

laya-onnx

Mattepiu

A ONNX build of laya, for running typed decisions on runtimes without PyTorch instead of a hosted API.

convaiinnovations/laya · apache-2.0

Community card

laya-onnx-fp16

sevenreasons

A ONNX build of laya, for running typed decisions on runtimes without PyTorch instead of a hosted API.

convaiinnovations/laya · apache-2.0

Browse another way

How to run ONNX

Step by step, with the runtime to install and what to watch for.

Every model

The full directory, with every facet listed in one place.

What is a decision model?

Start here if the category is new to you.

Find out what your data actually supports

We turn your historical decisions into a frozen test set, then tell you which model, which threshold, and how much of it you can safely automate.

Request a benchmark
Request a benchmark

New benchmark results, methodology notes and model additions. No product marketing.

Email us to be added
[email protected]
Models
All modelsGGUF buildsApple Silicon (MLX)Under 500M
Compare
All comparisonsOpen alternatives to JevDecision models vs LLMs
Built with
All projectsDeveloper toolingAI agentsBrowser
Self-host
Compatible serversllama.cppApple SiliconONNX
Browse
Commercial useRestricted licenceQuantised buildsBeyond English
Learn
What is a decision model?Decision models vs LLMsJev benchmark resultsHow to automate a decisionSet a confidence threshold
Company
Request a benchmark

© 2026 DecisionEval

Public benchmark results on this site are produced on frozen, documented test sets, with the method stated on each model page. They describe performance on those specific tasks only and are not a guarantee of performance on your own data. Vendor-reported figures are labelled as such and are not independently verified unless stated.