ModelsDeepSeek V4.1 Flash
DeepSeek V4.1 Flash
deepseek-ai/DeepSeek-V4.1-FlashDeepSeek V4.1 Flash is a sparse mixture-of-experts model from DeepSeek, and the first built on the company's Causal Encoder-Decoder (CED) architecture. It activates 8B parameters on input and 16B on output from a 552B-parameter backbone, an asymmetric split that keeps per-token compute low relative to the model's total size
VisionThinkingToolsRAG
✨New
Context
128K
Speed
Fast
Capabilities
Chat Completions
VisionYes
ThinkingYes
ToolsYes
InternetNo
RAGYes
InputText
OutputText
Parameters
| Parameter | Values | Recommended | |
|---|---|---|---|
| Temperature | min0 | max1 | 0.8 |
| Top P | min0 | max1 | 0.9 |
| Top K | min1 | max100 | 40 |
| Freq. Penalty | min-2 | max2 | 0 |
| Pres. Penalty | min-2 | max2 | 0 |
| Rep. Penalty | min1 | max2 | 1 |
| Max Tokens | min1 | max2048 | 1024 |
Credit Cost per Plan (8K tokens context)
Cost scales with the context length you use.
Context Length
Showing cost for requests up to 8,192 tokens.
Free3 cr/req
≈ 166 req/month
Basic3 cr/req
≈ 333 req/month
Starter3 cr/req
≈ 666 req/month
Pro2 cr/req
≈ 1,500 req/month
Pro+1 cr/req
≈ 5,000 req/month
Max1 cr/req
≈ 12,000 req/month
Max+1 cr/req
≈ 25,000 req/month
Ultimate1 cr/req
≈ 50,000 req/month
Context Window
128K
131,072 tokens