What does oev decide?

It answers 3 kinds of question: choice picks one option from a list you give it; score rates the state on a scale you define; noul returns the probability that a condition holds. Every answer is a typed value with a probability, never free text.

Pull the weights with systemone pull divyanshu-dhruv/oev and load them with the runtime named in the manifest.

Can I use oev commercially?

It is published under apache-2.0. Read the licence before shipping; the model card may add terms of its own.

What is oev built on?

Its base model is microsoft/deberta-v3-base, a oev architecture. The Model tree section shows the whole chain.

How much can oev read at once?

The publisher has not recorded a context length. The model has 184M parameters.

Its publisher reports 77.0% decision accuracy on LocalLLaMA/typed-decisions test set (2,000 questions), oev-base-td5 checkpoint fine-tuned on its train split; the maker's own run. Suites differ between publishers, so compare models on your own labelled examples.

oev: open AI decision model by Divyanshu Dhruv

OEV (oev-typed)

A 184M-parameter decision model on DeBERTa-v3-base by Divyanshu Dhruv. It packs a state, a typed question and its options into one sequence and returns a probability per option for choice, yes/no and score questions in one forward pass. English only.

The repository holds 15 checkpoints, not one model. The maker recommends student-r3 as the general default; oev-base-td5 is a specialist fine-tuned on the typed-decisions train split, and there are Banking77, MNLI and ANLI specialists. Calibration differs by checkpoint, from ECE 0.02 (td5 with a fitted temperature) to 0.23 out of domain. The card's Jev and Laya figures are quoted from those makers' tables, not run side by side. Checkpoints are PyTorch .pt files that load only through the maker's oev pip package, which fetches the DeBERTa backbone at load time.

What it decides

choice — picks one option from a set
score — places the input on an ordered scale
noul — answers a yes/no question with one probability

At a glance


Parameters	184M
Base model	`microsoft/deberta-v3-base`
Maker	Divyanshu Dhruv
Released	2026-09-23
License	apache-2.0
Reported accuracy	77.0%
Reported latency	22.2 ms p50 per question on a Tesla T4

Get the weights

pip install systemonemodels
systemone pull divyanshu-dhruv/oev

The files are served from the maker's Hugging Face repository, divyanshudhruv/oev-typed, and verified against the checksums recorded here.

Playground

Build a request for divyanshu-dhruv/oev. The form and the JSON are the same object; the response panel shows the shape every System One server returns.

Input

StateThe context the questions are asked about. Up to the model's context length.

A yes/no question. The answer is one probability that the condition holds.

Instructions

Response

Placeholder values, in the shape a run returns.

safenoul7%

This model doesn't answer live here. Send this request to your own server for this model; every System One server speaks this shape.

Call it

curl https://api.systemonemodels.ai/v1/systemone \
  -H "Authorization: Bearer $SYSTEMONE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"model":"divyanshu-dhruv/oev","state":"Task: clean up inactive accounts before the quarterly report.\nProposed tool call: delete_rows(table=\"customers\", where=\"last_login < 2023-01-01\")\nContext: the customers table is production; there is no backup newer than Friday.","questions":{"safe":{"type":"noul","instructions":"Is this action safe to run without a human approving it first?"}}}'

Decision accuracy	77.0%
Calibration error (ECE)	0.094
Median latency	22.2 ms

Divyanshu Dhruv: oev

OEV (oev-typed)

What it decides

At a glance

Get the weights

Read more

Playground

Divyanshu Dhruv: oev

OEV (oev-typed)

What it decides

At a glance

Get the weights

Read more

Playground

Providers

Pricing

Performance

Model tree

Activity

Documentation

systemone.yaml

FAQ