DeepSeek logo

DeepSeek V4 Pro

DeepSeek · Released Apr 24, 2026

Open SourcecodingreasoningbudgetCompare this model

$0.435

Input price / MTok

#7 of 27

$0.87

Output price / MTok

#6 of 27

1M

Context window

#15 of 28

72

Tokens / sec

#16 of 21

90.1

GPQA Diamond

#6 of 19

80.6

SWE-bench Verified

#3 of 12

93.5

LiveCodeBench

#1 of 8

Summary

DeepSeek V4 Pro is a open-source language model from DeepSeek, released on Apr 24, 2026.

At $0.435 input and $0.87 output per million tokens, it ranks #7 of 27 for input price, placing it mid-range on input cost.

Its context window of 1M tokens ranks #15 of 28 for context depth in this dataset.

It targets coding, reasoning, budget workloads.

Throughput is measured at 72 tokens per second (ranked #16 of 21 for throughput).

1.6T total params / 49B active MoE; supports Thinking and Non-Thinking modes; OpenAI + Anthropic API compatible.

Benchmarks

BenchmarkScoreMax%RankSource
GPQA Diamond90.110090.1%#6/19DeepSeek (self-reported in technical rep...
SWE-bench Verified80.610080.6%#3/12DeepSeek (self-reported in technical rep...
LiveCodeBench93.510093.5%#1/8DeepSeek (self-reported in technical rep...

Similar Models

Compare all 6 models

Models with the most shared benchmark coverage and closest scores to DeepSeek V4 Pro.

DeepSeek V4 Pro
DeepSeekDeepSeek V4 Flash
GoogleGemini 2.5 Pro
GoogleGemini 2.5 Flash
AnthropicClaude Sonnet 4.6
TencentTencent Hy3
GPQA Diamond#6 of 19
Graduate-level science reasoning
SWE-bench Verified#3 of 12
Real-world GitHub issue resolution
LiveCodeBench#1 of 8
Live competitive programming

Compare DeepSeek V4 Pro side by side

Pre-filled with 5 similar models

Open comparison

Speed source: https://artificialanalysis.ai/models/deepseek-v4-pro