ThinkingLevelSupportEN

Overview

• Some models support both thinking mode and normal mode
• You can switch by setting the Thinking Level in the API Settings
• Setting different thinking levels (OFF/LOW/MEDIUM/HIGH) will affect the model's thinking behavior, response time, and Token consumption

API Settings

• In Model Management -> Basic Settings, you can find the Thinking Level configuration item
  • OFF: Disable the thinking feature and use the standard response mode
  • LOW: Enable thinking and allocate a smaller Token budget, suitable for simple logic or saving consumption
  • MEDIUM: Allocate a medium Token budget, balancing translation quality and speed
  • HIGH: Allocate a higher Token budget, suitable for complex semantic understanding and rare topics
• The Model ID must be set correctly for the thinking level configuration to take effect

Supported Models

Model	OFF	LOW	MEDIUM	HIGH
GLM series	Off	On	On	On
Kimi series	Off	On	On	On
DeepSeek series	Off	On	On	On
Mimo-v2 series	Off	On	On	On
Qwen-3.5 series	Off	On	On	On
GPT-5 series	none	low	medium	high
Doubao series	minimal	low	medium	high
Gemini 2.5 Pro	128 Tokens	384 Tokens	768 Tokens	1024 Tokens
Gemini 2.5 Flash	0 Tokens	384 Tokens	768 Tokens	1024 Tokens
Gemini 2.5 Flash Lite	0 Tokens	512 Tokens	768 Tokens	1024 Tokens
Gemini 3 Pro	Low	Low	Low	High
Gemini 3.1 Pro	Low	Low	Medium	High
Gemini 3 Flash / Gemini 3.1 Flash	minimal	low	medium	high
Claude Sonnet 3.7 / Claude Haiku 4.x / Claude Sonnet 4.x / Claude Opus 4.x	Off	1024 Tokens	1536 Tokens	2048 Tokens