DeepSeek V4 Pro

by DeepSeek · chat

DeepSeek V4 Pro is the flagship DeepSeek V4 model for complex reasoning, coding, analysis, and agent tasks. It supports thinking and non-thinking modes, 1M context, up to 384K output, JSON output, tool calls, chat prefix completion, and FIM completion in non-thinking mode.

Pricing

Input: .6878/M tokens · Output: $3.3756/M tokens

Capabilities

Function Calling, JSON Mode, Streaming, Reasoning

Context: 1000K tokens

Max output: 384K tokens

Routes: 1/1 healthy

Performance

TTFT: 3080ms · Latency: 6144ms · Throughput: 76.1 tok/s