What does fern 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 reoring/fern and load them with the runtime named in the manifest.

Can I use fern commercially?

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

What is fern built on?

Its base model is qwen/qwen3.5-4b, a fern architecture. The Model tree section shows the whole chain.

How much can fern read at once?

Up to 1,024 tokens of state per request. The model has 4.0B parameters.

How accurate is fern?

Its publisher reports 42.0% decision accuracy on MMLU-Pro (2,000 questions, 10-way), accuracy against gold, the maker's own run of v3b. Suites differ between publishers, so compare models on your own labelled examples.

fern: open AI decision model by reoring

fern (v3b)

4B decision model distilled from DeepSeek-V4-Flash into Qwen3.5-4B. Send a state and up to 50 choice, score or noul questions; it returns a probability per option from one forward pass, with no generated text. Jev-compatible API, Apache-2.0 weights.

An independent project that copies the Jev request and response shape (fern serve, /v1/systemone). Up to 26 options per choice and 10 levels per score; prompts over 1,024 tokens lose the start of the state. Best on short states with clear criteria; weak at multi-step reasoning (MMLU-Pro accuracy 0.42 against gold). Mostly English. Most of the maker's figures are agreement with the teacher model, not accuracy against gold labels, and outputs are not calibrated against gold. The maker notes that training data includes Yelp reviews under non-commercial terms and asks users to check them before commercial use.

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	4B
Base model	`Qwen/Qwen3.5-4B`
Maker	reoring
Released	2026-09-25
License	apache-2.0
Reported accuracy	42.0%

Get the weights

pip install systemonemodels
systemone pull reoring/fern

The files are served from the maker's Hugging Face repository, reoring/fern, and verified against the checksums recorded here.

Playground

Build a request for reoring/fern. 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":"reoring/fern","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?"}}}'

reoring: fern

fern (v3b)

What it decides

At a glance

Get the weights

Read more

Playground

reoring: fern

fern (v3b)

What it decides

At a glance

Get the weights

Read more

Playground

Providers

Pricing

Performance

Model tree

Activity

Documentation

systemone.yaml

FAQ