# Zhipu AI

> Part of the ChinaAPI documentation. HTML: https://dash.chinaapi.ai/docs/model-makers/zhipu/

Every Zhipu AI model in the catalog, grouped by family: the request formats each one accepts, a request you can send, and what the gateway does with each field. Back to [all model makers](/docs/model-makers.md).

- **Families**: GLM · CogView · CogVideoX
- **Models**: 12
- **Tasks**: LLM · Image · Video · ASR · TTS
- **Request formats**: OpenAI-compatible · Anthropic Messages · Images API · Video API · Speech API · Transcription API
- **Capability profiles**: 9 / 12

## Models

Request formats are declared per model. A format listed for the maker does not mean every one of its models accepts it.

### GLM

| Model | Tasks | Request formats | Context | Capability profile | Evidence |
|---|---|---|---|---|---|
| `glm-5` | LLM | OpenAI-compatible · Anthropic Messages | 200K | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5-turbo` | LLM | OpenAI-compatible · Anthropic Messages | 200K | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5.1` | LLM | OpenAI-compatible · Anthropic Messages | 200K | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5.2` | LLM | OpenAI-compatible · Anthropic Messages | 1M | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5.3` | LLM | OpenAI-compatible · Anthropic Messages | 1M | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5.3-flash` | LLM | OpenAI-compatible · Anthropic Messages | 1M | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-5v-turbo` | LLM | OpenAI-compatible · Anthropic Messages | 200K | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-asr-2512` | ASR | Transcription API | — | [View profile](/docs/model-parameters.md) | Catalogue-declared |
| `glm-image` | Image | OpenAI-compatible · Images API | — | Not audited | — |
| `glm-tts` | TTS | Speech API | — | [View profile](/docs/model-parameters.md) | Live-verified |

### CogView

| Model | Tasks | Request formats | Context | Capability profile | Evidence |
|---|---|---|---|---|---|
| `cogview-4` | Image | OpenAI-compatible · Images API | — | Not audited | — |

### CogVideoX

| Model | Tasks | Request formats | Context | Capability profile | Evidence |
|---|---|---|---|---|---|
| `cogvideox-3` | Video | Video API | — | Not audited | — |

## Request formats

One key and one Base URL serve every format, so send the one your client already speaks. A format that no model from this maker declares says so, instead of showing a request that would fail.

**OpenAI-compatible**: `POST /v1/chat/completions`

```
curl https://api.chinaapi.ai/v1/chat/completions \
  -H "Authorization: Bearer $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'
```

**Anthropic Messages**: `POST /v1/messages`

```
curl https://api.chinaapi.ai/v1/messages \
  -H "x-api-key: $CHINAAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3",
    "max_tokens": 1024,
    "messages": [{ "role": "user", "content": "Hello" }]
  }'
```

**OpenAI Responses**: No model from this maker declares this request format in the catalog.

**Gemini generateContent**: No model from this maker declares this request format in the catalog.

## What the gateway does with each field

From the published capability manifest, one table per capability profile. A row covers every model in the profile; where the models differ, the row names each one and how it handles the field.

- **Native**: forwarded unchanged.
- **Translated**: carried across to the field the model's own format uses.
- **Normalized**: rewritten before the request leaves the gateway.
- **Model-specific**: the models in the profile handle the field differently, or only some of them accept it.
- **Partial**: checked on some models in the profile, not yet on the others.
- **Accepted but ignored**: taken without an error, and not used.
- **Not supported**: the gateway does not carry the field for that model.
- **Not verified**: nobody has checked yet. It is not a claim that the field works.
- **Not audited**: the model has no published capability profile, so nothing is stated about its fields.

### Capability profile `glm-text`

Models: `glm-5`, `glm-5-turbo`, `glm-5.1`, `glm-5.2`, `glm-5.3`, `glm-5.3-flash`, `glm-5v-turbo`

#### Capabilities

| Capability | Support | Evidence | Models |
|---|---|---|---|
| `prompt_cache` | Not verified | Catalogue-declared | All (7) |
| `reasoning` | Not verified | Catalogue-declared | All (7) |
| `streaming` | Native | Code-checked | All (7) |
| `structured_output` | Not verified | Catalogue-declared | All (7) |
| `tools` | Not verified | Catalogue-declared | All (7) |

#### Parameters

| Field | Type | Support | Evidence | Models |
|---|---|---|---|---|
| `max_completion_tokens` | integer | Translated | Code-checked | All (7) |
| `max_tokens` | integer | Not verified | Catalogue-declared | All (7) |
| `messages` | array | Native | Code-checked | All (7) |
| `model` | string | Native | Code-checked | All (7) |
| `stream` | boolean | Native | Code-checked | All (7) |
| `temperature` | number | Not verified | Catalogue-declared | All (7) |
| `top_p` | number | Normalized | Code-checked | All (7) |

### Capability profile `glm-speech`

Models: `glm-asr-2512`, `glm-tts`

#### Parameters

| Field | Type | Support | Evidence | Models |
|---|---|---|---|---|
| `file` | file | Model-specific | Live-verified | `glm-asr-2512`: Native · Live-verified |
| `input` | string | Model-specific | Live-verified | `glm-tts`: Native · Live-verified |
| `model` | string | Native | Live-verified | All (2) |
| `response_format` | string | Partial | Catalogue-declared | `glm-asr-2512`: Not verified · Catalogue-declared; `glm-tts`: Normalized · Live-verified |
| `voice` | string | Model-specific | Live-verified | `glm-tts`: Native · Live-verified |

## Evidence levels

A row, like a model, is shown at the weakest level of evidence among the fields it covers, never above it.

- **Catalogue-declared**: stated in the catalog; the gateway's handling has not been checked.
- **Code-checked**: the gateway's request code was checked to send it as stated.
- **Live-verified**: confirmed with a real request.

Each model's full profile, with its notes and value ranges, is on [Model parameters & capabilities](/docs/model-parameters.md).