> ## Documentation Index
> Fetch the complete documentation index at: https://narev.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# CreateDeepSeekV3Middleware

[**@ai-billing/deepseek**](../index)

***

[@ai-billing/deepseek](/docs/sdk/ai-billing/reference/deepseek/typedoc/index) / createDeepSeekV3Middleware

# Function: createDeepSeekV3Middleware()

> **createDeepSeekV3Middleware**\<`TTags`>(`options`): `LanguageModelV3Middleware`

Defined in: [deepseek/src/ai-sdk/language-model-middleware/v3/language-model-v3-deepseek-billing-middleware.ts:95](https://github.com/narevai/ai-billing/blob/main/packages/deepseek/src/ai-sdk/language-model-middleware/v3/language-model-v3-deepseek-billing-middleware.ts#L95)

Creates a V3 billing middleware for the DeepSeek provider (`@ai-sdk/deepseek`).
Derives token usage from DeepSeek's raw usage payload and resolves cost from pricing plus usage.

DeepSeek uses `prompt_cache_hit_tokens` / `prompt_cache_miss_tokens` for cache accounting, and
`completion_tokens_details.reasoning_tokens` for deepseek-reasoner models.

## Type Parameters

### TTags

`TTags` *extends* `JSONObject`

The shape of the tags object, extending DefaultTags.

## Parameters

### options

[`DeepSeekV3MiddlewareOptions`](/docs/sdk/ai-billing/reference/deepseek/typedoc/interfaces/DeepSeekV3MiddlewareOptions)\<`TTags`>

Billing options; see [DeepSeekV3MiddlewareOptions](/docs/sdk/ai-billing/reference/deepseek/typedoc/interfaces/DeepSeekV3MiddlewareOptions). A `priceResolver` is required.

## Returns

`LanguageModelV3Middleware`

A V3 billing middleware instance for DeepSeek.

## Example

```ts theme={null}
import { createDeepSeek } from '@ai-sdk/deepseek';
import { wrapLanguageModel } from 'ai';
import { createDeepSeekMiddleware } from '@ai-billing/deepseek';
import {
  consoleDestination,
  createObjectPriceResolver,
  type ModelPricing,
} from '@ai-billing/core';

const deepseek = createDeepSeek({ apiKey: process.env.DEEPSEEK_API_KEY });

const customPricingMap: Record<string, ModelPricing> = {
  'deepseek-chat': {
    promptTokens: 0.27 / 1_000_000,
    completionTokens: 1.10 / 1_000_000,
    inputCacheReadTokens: 0.07 / 1_000_000,
    inputCacheWriteTokens: 0,
  },
  'deepseek-reasoner': {
    promptTokens: 0.55 / 1_000_000,
    completionTokens: 2.19 / 1_000_000,
    inputCacheReadTokens: 0.14 / 1_000_000,
    inputCacheWriteTokens: 0,
    internalReasoningTokens: 2.19 / 1_000_000,
  },
};

const priceResolver = createObjectPriceResolver(customPricingMap);

const billingMiddleware = createDeepSeekMiddleware({
  destinations: [consoleDestination()],
  priceResolver,
});

const wrappedModel = wrapLanguageModel({
  model: deepseek('deepseek-chat'),
  middleware: billingMiddleware,
});
```