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

# DeepSeek V4 Pro

> DeepSeek's flagship V4 — strongest code generation in the open-weight tier, with reasoning chain support.

**Vendor:** DeepSeek
**Model ID:** `deepseek-v4-pro`
**Capability:** 64K context · tool use · streaming · structured output · reasoning
**Pricing:** per-token, pro tier ([live rate](https://bytespike.ai/pricing#text))

DeepSeek V4 Pro is the model to beat in the open-weight tier on code
generation tasks. Reach for it when the problem is structured —
implementations of well-defined algorithms, refactors with clear
constraints, codebase-style migrations — and the answer either
compiles or it doesn't. For freer-form tasks (architecture design,
prose, multi-step planning), GPT-5.5 and Claude Opus 4.8 still have an
edge.

## Request

```bash theme={null}
curl https://llm.bytespike.ai/v1/chat/completions \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [{"role": "user", "content": "Implement a thread-safe LRU cache in Rust."}]
  }'
```

### Body parameters

| Field             | Type    | Required | Default   | Notes                                                         |
| ----------------- | ------- | -------- | --------- | ------------------------------------------------------------- |
| `model`           | string  | yes      | —         | `deepseek-v4-pro`                                             |
| `messages`        | array   | yes      | —         | —                                                             |
| `max_tokens`      | integer | no       | model max | Max: 16384.                                                   |
| `temperature`     | number  | no       | 1.0       | —                                                             |
| `tools`           | array   | no       | —         | Function calling supported (parallel).                        |
| `response_format` | object  | no       | —         | JSON / structured output.                                     |
| `reasoning`       | object  | no       | —         | Optional reasoning chain — set `{"enabled": true}` to enable. |
| `stream`          | boolean | no       | false     | SSE streaming.                                                |

## Response

```json theme={null}
{
  "id": "chatcmpl-…",
  "object": "chat.completion",
  "model": "deepseek-v4-pro",
  "choices": [{"index": 0, "message": {"role": "assistant", "content": "use std::sync::..."}, "finish_reason": "stop"}],
  "usage": {"prompt_tokens": 32, "completion_tokens": 1248, "total_tokens": 1280}
}
```

## Code examples

<CodeGroup>
  ```bash cURL theme={null}
  curl https://llm.bytespike.ai/v1/chat/completions \
    -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
    -H "content-type: application/json" \
    -d '{"model": "deepseek-v4-pro", "messages": [{"role": "user", "content": "Implement a thread-safe LRU cache in Rust."}]}'
  ```

  ```python Python theme={null}
  from openai import OpenAI

  client = OpenAI(base_url="https://llm.bytespike.ai/v1", api_key="$BYTESPIKE_API_KEY")
  resp = client.chat.completions.create(
      model="deepseek-v4-pro",
      messages=[{"role": "user", "content": "Implement a thread-safe LRU cache in Rust."}],
  )
  print(resp.choices[0].message.content)
  ```

  ```javascript Node theme={null}
  import OpenAI from "openai"

  const client = new OpenAI({
    baseURL: "https://llm.bytespike.ai/v1",
    apiKey: process.env.BYTESPIKE_API_KEY,
  })
  const resp = await client.chat.completions.create({
    model: "deepseek-v4-pro",
    messages: [
      { role: "user", content: "Implement a thread-safe LRU cache in Rust." },
    ],
  })
  console.log(resp.choices[0].message.content)
  ```
</CodeGroup>

## Streaming + caching

`"stream": true` for SSE. Automatic prompt caching on stable prefixes.

## Errors

| Code                        | Trigger  | Billed?         |
| --------------------------- | -------- | --------------- |
| 400 / 401 / 402 / 422 / 429 | Standard | No              |
| 5xx                         | Upstream | No (auto-retry) |

## When to use

* Structured code generation in well-defined languages.
* Algorithm implementations, refactors with hard constraints.
* For lower latency, see [DeepSeek V4 Flash](/api-reference/text/deepseek-v4-flash).
* For multi-step planning beyond code, see [GPT-5.5](/api-reference/text/gpt-5-5) or [Claude Opus 4.7](/api-reference/text/claude-opus-4-7) (now superseded by Opus 4.8).

## Limits

| Limit                    | Value          |
| ------------------------ | -------------- |
| Context window           | 64K tokens     |
| Max output               | 16384 tokens   |
| Supports tool use        | Yes (parallel) |
| Supports vision          | No             |
| Supports streaming       | Yes            |
| Supports prompt caching  | Automatic      |
| Supports reasoning chain | Yes            |