DeepSeek V4 Flash

DeepSeek's V4-generation efficiency model: 284B-parameter MoE with 13B activated, 1M token context, optimized for fast inference at ultra-low cost. DeepSeek Fast Low-Cost V4, with a 1M token context window.

POST/v1/ai/chat

$0.01/call

Overview

Property	Value
Model ID	`deepseek/deepseek-v4-flash`
Context Window	1,048,576 tokens
Max Output	384,000 tokens
Input Price	$0.13 / 1M tokens
Output Price	$0.27 / 1M tokens

Usage

const res = await fetch('https://api.yepapi.com/v1/ai/chat', {
  method: 'POST',
  headers: {
    'x-api-key': 'YOUR_API_KEY',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({
    model: 'deepseek/deepseek-v4-flash',
    messages: [{ role: 'user', content: "Summarize the key differences between mixture-of-experts and dense transformer models, and when each is the right choice." }],
  }),
});
const { data } = await res.json();
console.log(data.message.content);

curl -X POST https://api.yepapi.com/v1/ai/chat \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek/deepseek-v4-flash", "messages": [{"role": "user", "content": "Summarize the key differences between mixture-of-experts and dense transformer models, and when each is the right choice."}]}'

Request Body

Parameter	Type	Required	Description	Default
`model`	`string`	Yes	Model ID (e.g. `deepseek/deepseek-v4-flash`)	—
`messages`	`Message[]`	Yes	Array of `{ role, content }` objects	—
`maxTokens`	`number`	No	Maximum tokens in the response	Model default
`temperature`	`number`	No	Sampling temperature (0.0–2.0)	`1.0`
`topP`	`number`	No	Nucleus sampling threshold	`1.0`
`frequencyPenalty`	`number`	No	Penalize repeated tokens	`0`
`presencePenalty`	`number`	No	Penalize tokens already present	`0`
`stream`	`boolean`	No	Enable SSE streaming	`false`

Info

All AI models use the /v1/ai/chat endpoint. Specify the model with the model field.

Response

{
  "ok": true,
  "data": {
    "model": "deepseek/deepseek-v4-flash",
    "message": {
      "role": "assistant",
      "content": "A dense transformer runs every parameter for every token, so quality and cost scale together. A mixture-of-experts model routes each token to a few expert sub-networks, so it can hold far more parameters while computing only a fraction per token: cheaper and faster at the same quality, but harder to serve because all experts must stay in memory. Choose dense for small models and simple deployment, MoE for large models where inference cost matters."
    },
    "usage": {
      "promptTokens": 24,
      "completionTokens": 180,
      "totalTokens": 204
    }
  }
}

Streaming

Set "stream": true to receive Server-Sent Events. Each chunk contains a delta object:

data: {"delta":{"content":"A"},"model":"deepseek/deepseek-v4-flash","index":0}
data: [DONE]

Under the Hood

We handle auth, billing, and response normalization — you just send messages.