DeepSeek V4 Pro

DeepSeek V4 Pro is generally available on app, web, and API. Ready to experience the power of AI? Start your journey here!

Platform: Replicate

Language ModelReasoningTool CallingLong Context

DeepSeek API

Off-peak: $0.022/1M cache-hit input, $0.66/1M cache-miss input, $1.98/1M output; peak: $0.044, $1.32, and $3.96 respectively

Commercial

🚀Function Overview

DeepSeek's generally available Pro model for app, web, and API use. Ready to experience the power of AI? Start your journey here!

Key Features

1M token context length in the official DeepSeek model table
384K maximum output in the official DeepSeek model table
Low, high, and max reasoning effort plus non-thinking mode switching guidance
JSON output and tool calls
Chat prefix completion beta and FIM completion in non-thinking mode
OpenAI-format and Anthropic-format API base URLs
Responses API support in the current official model table
Stable `deepseek-v4-pro` request ID mapped to the hosted `DeepSeek-V4-Pro-0813` backend
Generally available on DeepSeek app, web, and API since August 13, 2026
Off-peak pricing of $0.022 cache-hit input, $0.66 cache-miss input, and $1.98 output per 1M tokens
Peak pricing of $0.044 cache-hit input, $1.32 cache-miss input, and $3.96 output per 1M tokens
No separately identified 0813 public checkpoint as of August 23, 2026

Use Cases

•Long-context chat and agent workflows
•Coding-agent sessions that need tool calls and cache-aware cost control
•Structured JSON output generation
•API experiments comparing DeepSeek pricing against other frontier or long-context models

⚙️Input Parameters

messages

array

Chat messages sent to the DeepSeek API using OpenAI-format or Anthropic-format compatible endpoints.

thinking_mode

string

DeepSeek documents support for both thinking and non-thinking modes, with thinking enabled by default.

tools

array

Optional tool definitions for tool-calling workflows.

response_format

object

Optional JSON output controls when structured output is needed.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "deepseek-v4-pro",
  "messages": [
    {
      "role": "user",
      "content": "Summarize the tradeoffs of cached input pricing for a long coding-agent session."
    }
  ],
  "thinking_mode": "default"
}

Output Results

A chat completion response from the DeepSeek API. Verify the current request and response schema in the official DeepSeek API documentation before production use.

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: DeepSeek API
Commercial Use: Supported
Pricing: Off-peak: $0.022/1M cache-hit input, $0.66/1M cache-miss input, $1.98/1M output; peak: $0.044, $1.32, and $3.96 respectively
Platform: Replicate

Related Keywords

DeepSeek V4 ProDeepSeek-V4-Pro-0813deepseek-v4-proDeepSeek Responses APIDeepSeek API pricing1M context modeltool calling modelcached input pricingthinking mode

Related Models

DeepSeek V4 Flash

DeepSeek coding-agent model with a public-beta DeepSeek-V4-Flash-0731 API version, official MIT-licensed weights, thinking modes, 1M context, 384K maximum output, tool calls, Responses API support, and Codex integration.

Bielik 1.5B v3 Instruct

Bielik-1.5B-v3-Instruct is a generative text model featuring 1.6 billion parameters. It is result of collaboration between the open-science/open-souce project SpeakLeash and the High Performance Computing (HPC)

Cordia-A6 Text Generation Model

A model for generating text sequences based on input prompts and adjustable parameters.