# wfzyx: von

> A ModernBERT-large encoder with an option-marker head. Premise and options are packed into one sequence and each option's marker is scored in a single bidirectional pass; version 1.2 makes the scoring order-invariant.

- Page: https://systemonemodels.ai/wfzyx/von
- API: https://api.systemonemodels.ai/v1/models/wfzyx/von
- Download: `pip install systemonemodels && systemone pull wfzyx/von`

## Facts

| | |
|---|---|
| Maker | wfzyx (https://systemonemodels.ai/wfzyx) |
| Decides | choice, score, noul, classify |
| Architecture | von |
| Base model | answerdotai/modernbert-large |
| Parameters | 395M |
| Context | 8K tokens |
| Licence | apache-2.0 |
| Availability | Open weights |
| Released | 2026-09-19 |
| Latest version | 1.2.0 |

## Reported evaluation

Suite: JevBench public standard tier. Numbers are the publisher's own.

- Decision accuracy: 63.9%
- Calibration error (ECE): 0.045
- Median latency: 18 ms

## Model card

<!-- generated by scripts/seed_catalog.py; edit content/models/catalog.yaml -->

# Von 1.2

A ModernBERT-large encoder with an option-marker head. Premise and options are packed into one sequence and each option's marker is scored in a single bidirectional pass; version 1.2 makes the scoring order-invariant.

On JevBench's public standard tier Von scores 63.9% (easy 100%, hard 36.9%) with a calibration error of 0.045, though the calibration map was fitted on the same 231 items. Context is 8,192 tokens, and von-sdk (pip and npm) serves /v1/systemone. Latency is about 18 ms on a GPU.

## 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 calibrated probability
- **classify** — assigns a category from a fixed taxonomy

## At a glance

| | |
|---|---|
| Parameters | 395M |
| Base model | `answerdotai/modernbert-large` |
| Maker | wfzyx |
| Released | 2026-09-19 |
| License | apache-2.0 |
| Reported accuracy | 63.9% |
| Reported latency | ~18 ms on a GPU |

## Get the weights

```bash
pip install systemonemodels
systemone pull wfzyx/von
```

The files are served from the maker's Hugging Face repository, [`wfzyx/von`](https://huggingface.co/wfzyx/von), and verified against the checksums recorded here.

## Read more

- [Source](https://github.com/wfzyx/von)
- [Weights on Hugging Face](https://huggingface.co/wfzyx/von)

---

*This page was opened by System One for wfzyx, who can claim the organisation and take it over at any time.*

---

From System One Models — https://systemonemodels.ai/ · every System One model: https://systemonemodels.ai/system-one-models