Supported Models and ParametersText Models
Text Models
Review shared text-model fields and open the single-model page for exact guidance.
Text models are different from image and video task models. They usually return text generation results directly instead of relying on async task polling.
The following list reflects currently valid public routes in the China-region live configuration. Actual availability, region, API-key access, and pricing always follow GET /v1/models.
Shared Text Fields
| Field | Type | Required | Description |
|---|---|---|---|
model | string | Yes | Target text model ID |
input | string | Yes | User input content |
messages | array | Yes (Chat) | Multi-turn system, user, and assistant messages |
instructions | string | No (Responses) | System-level instructions for Responses |
stream | boolean | No | Whether to stream the response |
reasoning | object | Schema-dependent | Send only when declared by schema.fields |
Current Text Models
| Model | Region | Best For | Details |
|---|---|---|---|
gpt_5_4_mini | CN | Lightweight, cost-first | Public |
gpt_5_5 | Live config | General text generation | Public |
gpt_5_6_luna | Live config | Text generation | Public |
gpt_5_6_sol | Live config | Text generation | Public |
gpt_5_6_terra | Live config | Text generation | Public |
gpt_6_astra | CN | Next-generation general text | Public |
gpt_6_sol | CN | Next-generation cost-efficient text | Public |
gpt_6_luna | CN | Next-generation ultra-low-cost text | Public |
glm_5_3_flash | CN | Fast text generation | Public |
glm_5_3 | CN | Quality-first text generation | Public |
claude_fable_5 / claude_fable_5_1 | All | General text generation | Public |
claude_haiku_4_5 | All | Fast, cost-efficient generation | Public |
claude_opus_4_7 / claude_opus_4_8 / claude_opus_5 | All | High-quality complex tasks | Public |
claude_sonnet_4_6 / claude_sonnet_5 | All | Balanced quality and speed | Public |
gemini_3_1_pro | All | High-quality general text | Public |
gemini_3_5_flash / gemini_3_6_flash / gemini_3_7_flash / gemini_3_8_flash | All | Fast text generation | Public |
grok_4_5 / grok_4_6 | All | General text generation | Public |
deepseek_v4_flash / deepseek_v4_pro | CN | Fast or quality-first generation | Public |
deepseek_v4_flash_vision_exp | CN | Image understanding and text generation | Public |
Why This Section Exists Now
- Text models support system instructions, multi-turn history, and streaming; advanced fields such as reasoning are model-specific and must follow the live schema.
- Use
messages[].role = systemfor Chat Completions and top-levelinstructionsfor Responses. Do not send a customsystemPromptfield. - Use
/v1/chat/completionsor/v1/responsesfor text models, not the media/v1/generationsendpoint.