# DeepSeek-V4-Pro

> DeepSeek's large-scale open-weights mixture-of-experts model for advanced reasoning and coding workloads.

## Key facts

- Model author: DeepSeek
- Input: $1.2686 per 1M tokens (starting price)
- Output: $2.5372 per 1M tokens (starting price)
- Cached input: $0.044 per 1M tokens
- Context window: 1,000,000 tokens
- Maximum output: 384,000 tokens
- Input modalities: text
- Output modalities: text
- Capabilities: reasoning, tools
- Released: 2026-08-13
- Model IDs: `baidu/deepseek-v4-pro`, `deepseek/deepseek-v4-pro`, `jd/deepseek-v4-pro`
- Model page: https://www.realrelay.ai/model/deepseek/deepseek-v4-pro

## Overview

DeepSeek-V4-Pro is a large-scale mixture-of-experts model from DeepSeek with open weights, designed for advanced reasoning and coding. It supports both thinking and non-thinking modes, with thinking as the default, plus tool calling and JSON output. It takes text input, returns text, and offers a 1M-token context window.

DeepSeek-V4-Pro has a 1M-token context window and can generate up to 384K output tokens. It accepts text input and returns text output. It was released on August 13, 2026. Its weights are open.

## Capabilities

DeepSeek-V4-Pro supports reasoning and tool calling. It accepts text as input and produces text as output.

Thinking mode lets DeepSeek-V4-Pro work through hard problems step by step, while non-thinking mode returns direct answers. Tool calling and JSON output make it practical for agents and pipelines that need machine-readable results.

## Pricing

DeepSeek-V4-Pro starts from $1.27 per 1M input tokens and from $2.54 per 1M output tokens. Cache reads start from $0.044 per 1M tokens.

Pricing is tiered, and the price shown is the starting tier. DeepSeek-V4-Pro is available through RealRelay's OpenAI-compatible API.

## Use cases

Use DeepSeek-V4-Pro for reasoning-heavy coding tasks, such as debugging a difficult issue or planning a refactor, where the 1M-token window holds the relevant code and thinking mode works through the problem.

With JSON output and tool calling, it can also drive data-extraction and agent pipelines that must return structured results from long text inputs.

## Call it

OpenAI-compatible chat completions:

```bash
curl https://www.realrelay.ai/v1/chat/completions \
  -H "Authorization: Bearer $REALRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek/deepseek-v4-pro", "messages": [{"role": "user", "content": "Hello"}]}'
```

## FAQ

### What is DeepSeek-V4-Pro?

DeepSeek-V4-Pro is a large-scale mixture-of-experts model from DeepSeek with open weights, designed for advanced reasoning and coding. It supports reasoning and tool calling and accepts text input.

### How much does DeepSeek-V4-Pro cost?

DeepSeek-V4-Pro starts from $1.27 per 1M input tokens and from $2.54 per 1M output tokens, with cache reads from $0.044 per 1M tokens. Pricing is tiered, and these are the starting-tier prices.

### What is the context window of DeepSeek-V4-Pro?

DeepSeek-V4-Pro has a 1M-token context window and can generate up to 384K tokens in a single response.

### What inputs and outputs does DeepSeek-V4-Pro support?

DeepSeek-V4-Pro accepts text input and returns text output.

### Does DeepSeek-V4-Pro support tool calling?

Yes. DeepSeek-V4-Pro supports tool calling, so it can call functions you define and use the results in multi-step tasks. It also supports reasoning.

### When was DeepSeek-V4-Pro released?

DeepSeek-V4-Pro was released on August 13, 2026.

### How do I call DeepSeek-V4-Pro via the API?

Call DeepSeek-V4-Pro through RealRelay's OpenAI-compatible API using the model ID deepseek/deepseek-v4-pro.