DeepSeek-V4-Flash-Vision 与 Qwen3.8-Flash-Next 思考模式控制说明
1. 参数对照
| 模型 | 关闭思考 | 开启思考 | 思考强度 |
|---|---|---|---|
| DeepSeek-V4-Flash-Vision-Exp | thinking: false | thinking: true | low / high / max |
| Qwen3.8-Flash-Next | enable_thinking: false | enable_thinking: true | low / medium / xhigh |
2. DeepSeek-V4-Flash-Vision
2.1 关闭思考
通过 chat_template_kwargs.thinking=false:
{
"chat_template_kwargs": {
"thinking": false
}
}完整请求示例:
curl http://127.0.0.1:8030/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "DeepSeek-V4-Flash-Vision",
"messages": [
{
"role": "user",
"content": "介绍一下 vLLM"
}
],
"chat_template_kwargs": {
"thinking": false
}
}'2.2 开启思考
{
"chat_template_kwargs": {
"thinking": true
}
}2.3 控制思考强度
DeepSeek-V4-Flash-Vision支持通过 reasoning_effort 控制思考强度。
当不指定思考强度时,默认为 low
Low
{
"chat_template_kwargs": {
"thinking": true,
"reasoning_effort": "low"
}
}High
{
"chat_template_kwargs": {
"thinking": true,
"reasoning_effort": "high"
}
}Max
{
"chat_template_kwargs": {
"thinking": true,
"reasoning_effort": "max"
}
}完整示例:
curl http://127.0.0.1:8030/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "DeepSeek-V4-Flash-Vision",
"messages": [
{
"role": "user",
"content": "分析 Transformer 和 Mamba 的主要区别"
}
],
"temperature": 1.0,
"top_p": 0.95,
"chat_template_kwargs": {
"thinking": true,
"reasoning_effort": "high"
}
}'3. Qwen3.8-Flash-Next
3.1 关闭思考
Qwen 使用 enable_thinking:
{
"chat_template_kwargs": {
"enable_thinking": false
}
}完整示例:
curl http://127.0.0.1:8040/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3.8-Flash-Next",
"messages": [
{
"role": "user",
"content": "介绍一下 vLLM"
}
],
"temperature": 0.7,
"top_p": 0.8,
"presence_penalty": 1.5,
"top_k": 20,
"chat_template_kwargs": {
"enable_thinking": false
},
"max_tokens": 1024
}'3.2 开启思考
{
"chat_template_kwargs": {
"enable_thinking": true
}
}3.3 控制思考强度
Qwen3.8-Flash-Next 支持三个强度档位:
lowmediumxhigh
注意:
- 没有
high - 没有
max - 最高档是
xhigh - 不指定时默认
xhigh
Low
{
"reasoning_effort": "low",
"chat_template_kwargs": {
"enable_thinking": true
}
}Medium
{
"reasoning_effort": "medium",
"chat_template_kwargs": {
"enable_thinking": true
}
}XHigh
{
"reasoning_effort": "xhigh",
"chat_template_kwargs": {
"enable_thinking": true
}
}完整示例:
curl http://127.0.0.1:8040/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "Qwen3.8-Flash-Next",
"messages": [
{
"role": "user",
"content": "设计一个高可用的 AI Gateway 架构"
}
],
"reasoning_effort": "medium",
"temperature": 1.0,
"top_p": 0.95,
"top_k": 20,
"presence_penalty": 0.0,
"chat_template_kwargs": {
"enable_thinking": true
}
}'4. 总结
最核心的区别可以记成:
DeepSeek V4
├── thinking
└── reasoning_effort
├── low
├── high
└── maxQwen3.8 Flash Next
├── enable_thinking
└── reasoning_effort
├── low
├── medium
└── xhigh
评论暂时无法加载。