> For the complete documentation index, see [llms.txt](https://docs.eyepop.ai/developer-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.eyepop.ai/developer-documentation/platform/abilities.md).

# Abilities

The vision tasks you name in a Pop — pretrained models, custom trained models, and vision-language abilities you define yourself

An **ability** is a vision task you name in a [Pop](https://docs.eyepop.ai/developer-documentation/platform/pop.md) — detect objects, classify a scene, read text — and a model is what actually performs the analysis behind it. EyePop offers two kinds of model, and they differ in who can use them.

**Pretrained models** ship with the platform. They live in the `eyepop.` namespace, cover the tasks most projects need — detection, pose, OCR, segmentation, finding things you describe in words — and are available to every account with no training and no setup.

**Custom trained models** are the ones you build from your own labeled data, for the cases no pretrained model covers. A custom model belongs to the account that trained it and is accessible only to that account — your data and the model derived from it stay yours. See [Train Your Own Model](https://docs.eyepop.ai/developer-documentation/train-your-own-model.md) for the full workflow.

Both kinds run the same way once they exist: an [inference component](https://docs.eyepop.ai/developer-documentation/platform/pop/pop-components.md) in a [Pop](https://docs.eyepop.ai/developer-documentation/platform/pop.md) names one, and the rest of the pipeline does not care which kind it got.

### Fixed classes and open vocabulary

Pretrained models answer in one of two ways.

**Fixed classes.** The labels are chosen when the model is trained. `eyepop.vehicle:latest` finds bicycles, cars, motorcycles, buses, trains and trucks, and nothing else: there is no way to ask it for a forklift. A fixed-class model is fast and predictable, and it is the right choice whenever its labels cover your question. See [Pretrained, fixed classes](https://docs.eyepop.ai/developer-documentation/platform/abilities/pretrained-fixed-classes.md) for the catalog.

**Open vocabulary.** The labels are chosen **per request**. You describe what to look for in words, in the component's `params.prompts`, and the model reports matches under the labels you give it:

```json
{
    "type": "inference",
    "ability": "eyepop.sam3.detect:latest",
    "params": { "prompts": [{ "prompt": "forklift" }, { "prompt": "pallet" }] }
}
```

The same ability finds forklifts in one Pop and shopping carts in the next, without training anything. Open-vocabulary models are larger and slower than fixed-class ones, so reach for them when no fixed label set covers the question, or to try an idea before training a custom model. See [Open vocabulary](https://docs.eyepop.ai/developer-documentation/platform/abilities/open-vocabulary.md) for the abilities, their parameters and worked examples; [Promptable VLM tasks](https://docs.eyepop.ai/developer-documentation/platform/abilities/open-vocabulary/promptable-vlm-tasks.md) covers the vision-language models among them.

### Abilities you define

You can also define an ability yourself on a vision-language model: a free-form prompt, a free-form prompt with a response schema, or a task with your own labels. Publishing it gives it an alias under your namespace. It runs on the VLM you choose: `qwen3`, EyePop's own and the default, or the remote `gemini_4_argon` or `gpt_6_astra`. See [User defined VLM tasks](https://docs.eyepop.ai/developer-documentation/platform/abilities/user-defined-vlm-tasks.md) and [Choosing a model](https://docs.eyepop.ai/developer-documentation/platform/abilities/user-defined-vlm-tasks.md#choosing-a-model).

### Referencing a model

Name a pretrained model by its alias and tag, exactly as you would an ability:

```json
{ "type": "inference", "ability": "eyepop.person:latest" }
```

A custom trained model is named by its uuid instead, or by the alias [publishing it as an ability](https://docs.eyepop.ai/developer-documentation/platform/abilities/user-defined-vlm-tasks.md) gives it under your own namespace:

```json
{ "type": "inference", "abilityUuid": "066ae8f1fc0174138000cb8bcfa7bdeb" }
```

For the full searchable catalog including community and custom abilities, see the [Abilities Hub](https://www.eyepop.ai/abilities). To create your own from a prompt, use the [dashboard](https://dashboard.eyepop.ai/dashboard).

### Sized variants

Several pretrained models come in sized variants, named by a suffix on the alias: `eyepop.person.large:latest`, `eyepop.person.nano:latest` and so on. A smaller variant trades slightly lower precision for higher throughput and lower latency. Variants of a detection model share its labels, so switching between them changes nothing else in the Pop; vectors from two sizes of an embedding model are not comparable. See [Sized variants](https://docs.eyepop.ai/developer-documentation/platform/abilities/pretrained-fixed-classes.md#sized-variants) for the details and the models that have them.

## Using a model

An inference component naming a model can also tag its output group, filter weak detections with `confidenceThreshold`, and pass parameters such as `prompts` to the abilities that accept them. Tracking optionally names a re-identification model such as `eyepop.person.reid:latest`.

The [Quickstart](https://docs.eyepop.ai/developer-documentation/quickstart.md) shows a minimal Pop-first prediction.

### Next steps

* [Pretrained, fixed classes](https://docs.eyepop.ai/developer-documentation/platform/abilities/pretrained-fixed-classes.md) — the catalog of models with a fixed label set
* [Open vocabulary](https://docs.eyepop.ai/developer-documentation/platform/abilities/open-vocabulary.md) — models that find and classify what you describe in words
* [Components](https://docs.eyepop.ai/developer-documentation/platform/pop/pop-components.md) — every component type and attribute
* [Examples](https://docs.eyepop.ai/developer-documentation/platform/pop/pop-examples.md) — worked pipelines end to end
* [Train Your Own Model](https://docs.eyepop.ai/developer-documentation/train-your-own-model.md) — build a custom model from your own data
* [User defined VLM tasks](https://docs.eyepop.ai/developer-documentation/platform/abilities/user-defined-vlm-tasks.md) — define your own ability with a prompt, a response schema or a task

Questions about building a Pop? Reach us at <help@eyepop.ai>.
