# Frequently Asked Questions

## How to View the List of Supported Models

You can obtain all supported models via the `/v1/models` endpoint:

```shell
curl https://api-us-ca.umodelverse.ai/v1/models \
  -H "Content-Type: application/json"
```

It is recommended to use the `Cherry Studio` client. After configuring the API address, click the management button to visually view all supported models. For detailed configuration tutorials, please refer to: [Cherry Studio Configuration Tutorial](/docs/modelverse/best_practice/scenario/cherry-studio).

<img src="https://cdn.udelivrs.com/2026/01/747aaa503b4047f61007131f8b7ba5a5_1767698734503.png" data-origin="https://cdn.udelivrs.com/2026/01/747aaa503b4047f61007131f8b7ba5a5_1767698734503.png" alt="model" class="medium-zoom-image" style="">

## GPT-5/o Series Model Parameter Error

<img src="https://cdn.udelivrs.com/2026/01/e2cff576b72cf2232ecb813793aeb86d_1767698734497.png" data-origin="https://cdn.udelivrs.com/2026/01/e2cff576b72cf2232ecb813793aeb86d_1767698734497.png" alt="gpt-5-para" class="medium-zoom-image" style="">

OpenAI has marked the `max_tokens` parameter as deprecated. The GPT-5 series and o series models no longer support this parameter. For more details, please refer to the [OpenAI Official Documentation](https://platform.openai.com/docs/api-reference/chat/create).

If you need to pass image files, please use base64 encoding. URL method is not supported yet.

## Explanation of gpt-5-codex and gpt-5.1-codex Series Models

These two series of models only support the `/v1/response/` endpoint.

## How to Enable/Disable the Thought Process for Gemini Series Models

Use the Gemini protocol's `v1beta/models` endpoint and configure the following parameters. For detailed documentation, please refer to [Gemini Protocol Compatibility](/docs/modelverse/modelverse/text_api/gemini_compatible):

```json
"generationConfig": {
    "thinkingConfig": {
        "include_thoughts": true,
        "thinkingBudget": 0
    }
}
```

Parameter Descriptions:
- `include_thoughts`: Set to `true` to enable thought process output, set to `false` to disable
- `thinkingBudget`: Control thought time budget

## Claude *Series Model Description*

Claude Sonnet 4.5 only supports specifying one of the `temperature` or `top_p` parameters, and both cannot be used simultaneously. For more information, see the [Official Documentation](https://docs.aws.amazon.com/zh_cn/bedrock/latest/userguide/model-parameters-anthropic-claude-messages-request-response.html).

## Security Level Description for Gemini Series Models

Gemini API provides flexible security setting features, allowing you to adjust content filtering policies based on the application scenario in both prototyping and production environments. By configuring the security level, you can precisely control the strictness of content review in five filter categories.

Related Documentation:
- [Gemini API Compatibility Documentation](/docs/modelverse/modelverse/text_api/gemini_compatible)
- [Google Official Security Settings Documentation](https://ai.google.dev/gemini-api/docs/safety-settings?hl=zh-cn)

### Security Setting Configuration Example

The following example shows how to disable all content filtering (`BLOCK_NONE`), suitable for development scenarios requiring maximum flexibility:

```json
{
  "safetySettings": [
    {
      "category": "HARM_CATEGORY_HATE_SPEECH",
      "threshold": "BLOCK_NONE"
    },
    {
      "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
      "threshold": "BLOCK_NONE"
    },
    {
      "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
      "threshold": "BLOCK_NONE"
    },
    {
      "category": "HARM_CATEGORY_HARASSMENT",
      "threshold": "BLOCK_NONE"
    },
    {
      "category": "HARM_CATEGORY_CIVIC_INTEGRITY",
      "threshold": "BLOCK_NONE"
    }
  ]
}
```

### Available Filter Categories

- `HARM_CATEGORY_HATE_SPEECH`: Hate speech
- `HARM_CATEGORY_SEXUALLY_EXPLICIT`: Sexually explicit content
- `HARM_CATEGORY_DANGEROUS_CONTENT`: Dangerous content
- `HARM_CATEGORY_HARASSMENT`: Harassment content
- `HARM_CATEGORY_CIVIC_INTEGRITY`: Civic integrity

### Available Threshold Options

The thresholds (HarmBlockThreshold) are used to block content when a specified harm probability is reached or exceeded:

- `HARM_BLOCK_THRESHOLD_UNSPECIFIED`: Unspecified threshold
- `BLOCK_LOW_AND_ABOVE`: Allows content with "negligible" risk, blocks content with "low" and above risk
- `BLOCK_MEDIUM_AND_ABOVE`: Allows content with "negligible" and "low" risk, blocks content with "medium" and above risk
- `BLOCK_ONLY_HIGH`: Allows content with "negligible", "low", and "medium" risk, only blocks content with "high" risk
- `BLOCK_NONE`: Allows all content, no blocking
- `OFF`: Turn off the security filter function

Choose appropriate threshold configurations based on your application requirements and usage scenarios. It is recommended to use stricter filtering strategies in production environments to ensure content safety.