---
title: "Models"
description: "The Model Catalog, model references and versions, and each model's properties."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.ezghcloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Models

The Model Catalog lists every model EZGH Cloud serves, in any product, with its versions and
properties. It's the same for everyone, and reading it needs no credential or permission.

## Catalog

| Model ID | Name | Task | Offered by | Latest version | License |
| --- | --- | --- | --- | --- | --- |
| `qwen3-embedding-0.6b` | Qwen3 Embedding 0.6B | Embeddings | Inference | `1.0.0` | Apache-2.0 |
| `unlimited-ocr` | Baidu Unlimited-OCR | OCR | OCR | `1.0.0` | MIT |

**Offered by** is where you can use a model. Models offered by Inference are called through the
OpenAI-compatible API ([Embeddings with OpenAI SDKs](/inference/embeddings)). Models offered by OCR
are used through [OCR processors](/ocr/models); the Inference API answers `404 model_not_found`
for them.

### qwen3-embedding-0.6b

[Qwen3-Embedding-0.6B](https://huggingface.co/Qwen/Qwen3-Embedding-0.6B) by Qwen: multilingual
text embeddings.

| Property | Value |
| --- | --- |
| Dimensions (default and maximum) | 1,024 |
| Minimum dimensions | 32 |
| Tokens per input | 8,192 |
| Inputs per request | 128 |
| Request size | 1 MiB |
| Instruction-aware | Yes |
| Multilingual | Yes |

| Version | Status | Released | Source |
| --- | --- | --- | --- |
| `1.0.0` | `ga` | 2026-10-10 | `Qwen/Qwen3-Embedding-0.6B` at `97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3` |

## Model references

A request names a model by its ID or a pinned version:

| Reference | Means |
| --- | --- |
| `qwen3-embedding-0.6b` | The model's latest version, which can change when a new version is released |
| `qwen3-embedding-0.6b:1.0.0` | Exactly that version |

IDs are lowercase and matched exactly. Every response names the version that ran, as
`<id>:<version>`. Pin a version when stored vectors must stay comparable with new ones.

## Versions and lifecycle

| Status | Served | Notes |
| --- | --- | --- |
| `preview` | Yes | |
| `ga` | Yes | |
| `deprecated` | Yes | Responses carry a `Deprecation` header, and a `Sunset` header with the retirement date once one is set |
| `retired` | No | `410 model_retired`. A retired version stays in the catalog |

## Reading the catalog

In the console, open **Inference**, then **Models**. With the CLI:

```sh
ezgh inference models list
ezgh inference models get qwen3-embedding-0.6b
```

Or read it from the API, with no credential:

```sh
curl https://inference-api.ezghcloud.com/public/v1/models
curl https://inference-api.ezghcloud.com/public/v1/models/qwen3-embedding-0.6b
```

The list is `{ "models": [...], "nextCursor": null }`, paged with `limit` (1 to 100, default 100)
and `cursor` (the previous page's `nextCursor`). Each model has `id`, `displayName`,
`description`, `publisher`, `license`, `task`, `modalities`, `offeredBy`, `latestVersion`,
`versions` (each with `version`, `status`, `releasedAt`, `deprecatedAt`, `retiresAt` and
`source`) and `properties`. Reads are limited per client address; over the limit is `429` with
`Retry-After`.

The OpenAI SDKs' `models.list()` and `models.retrieve()` work too, on the project's base URL
([list models](/inference-api/ListModels/), `inference.models.list`): they list the models
offered by Inference, with `owned_by` set to the publisher.

Source: https://docs.ezghcloud.com/inference/models/index.mdx