DeepSeek V4 Pro

Name: DeepSeek V4 Pro
Price: 1.32 USD
Author: DeepSeek

DeepSeek V4vPro 0813Current

DeepSeek(Privately Held)

Released August 13, 2026

API Docs Playground

OpenTools Scorenormalized /100

17.3

Valuescore / $M

OpenTools Score

Context1.0M tokens

Price (In / Out)$1.32/M / $3.96/M

CategoryLarge Language Model

Max Output393.2K tokens

About DeepSeek V4 Pro

DeepSeek V4 Pro 0813 is an open-weight Mixture-of-Experts reasoning model for long-context coding, tool use, and agentic workloads, with native Responses API support through the DeepSeek API.

Capabilities

reasoningcodingtool usestructured outputlong contextagenticresponses api

Input Modalities

text

Output Modalities

text

Technical Details

API Identifier: deepseek-v4-pro
Category: Large Language Model
Context Window: 1,048,576 tokens
Max Output Tokens: 393,216 tokens

Benchmarks

Performance scores for DeepSeek V4 Pro across standard benchmarks.

Artificial Analysis Intelligence Index v4.1Artificial Analysis · Aug 2026

44%

Artificial Analysis Intelligence Index v4.1.1Artificial Analysis · Aug 2026

53%

Artificial Analysis Intelligence Index v4.3Artificial Analysis

36.2829%

GPQA DiamondArtificial Analysis

92.8283%

SciCodeArtificial Analysis

51.0417%

Terminal-Bench 2.1Artificial Analysis

78.6517%

Artificial Analysis Intelligence Index v4.3.2Artificial Analysis

20.3605101726847%

Pricing

Token pricing for DeepSeek V4 Pro API usage.

Input Tokens

$1.32/M

per million tokens

Output Tokens

$3.96/M

per million tokens

Pricing Calculator

Input tokens/month

Output tokens/month

Input cost$1.32

Output cost$1.98

Estimated monthly cost$3.30

DeepSeek direct API, DeepSeek-V4-Pro-0813: headline prices are peak uncached input $1.32 and output $3.96 per million tokens. Off-peak input is $0.66 and output $1.98; cache-hit input is $0.044 peak/$0.022 off-peak. Peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday-Friday; all other hours are off-peak. DeepSeek has withdrawn the planned September 14, 2026 retirement and continues this API. Rates are conditional, not an unconditional off-peak offer.

Competing Models

Same pricing tier — direct alternatives to DeepSeek V4 Pro

Mid-Range

GLM 5.1 z-ai

26.2

$0.95/M / $3.15/MCompare

DeepSeek V3.2 Speciale deepseek

93.9

$0.40/M / $1.20/MCompare

MiniMax M2.7 minimax

57.2

$0.30/M / $1.20/MCompare

Trinity Large Thinking arcee-ai

52.5

$0.22/M / $0.85/MCompare

Hermes 4 405B nous-research

$1.00/M / $3.00/MCompare

Kimi K2 Thinking moonshot-ai

47.2

$0.60/M / $2.50/MCompare

WizardLM-2 8x22B microsoft

$0.62/M / $0.62/MCompare

GLM 5 Turbo z-ai

30.0

$1.20/M / $4.00/MCompare

MiniMax M2.5 minimax

101

$0.12/M / $0.99/MCompare

Kimi K2 0905 moonshot-ai

44.3

$0.40/M / $2.00/MCompare

GLM 5 z-ai

46.5

$0.72/M / $2.30/MCompare

DeepSeek V3.1 Terminus deepseek

82.6

$0.21/M / $0.79/MCompare

MiniMax M2-her minimax

$0.30/M / $1.20/MCompare

Kimi K2 0711 moonshot-ai

42.2

$0.57/M / $2.30/MCompare

GLM-5.2 z-ai

18.2

$1.40/M / $4.40/MCompare

GPT-5.6 Luna openai

70.7

$0.20/M / $1.20/MCompare

Gemini 3.5 Flash-Lite google

15.8

$0.30/M / $2.50/MCompare

Muse Spark 1.1 meta

19.9

$1.25/M / $4.25/MCompare

Nemotron 3.5 Lightning 30B A3B nvidia

21.1

$0.10/M / $0.95/MCompare

Gemini 3.7 Flash google

31.2

$0.75/M / $3.75/MCompare

Apodex 1.1 apodex

27.2

$0.24/M / $2.40/MCompare

DeepSeekPrivately Held

Unravel the mystery of AGI with curiosity.

1 Tools6 ModelsFounded 2023Hangzhou, China

View full profile

1 Tools·6 Models

Tools by DeepSeek

Other AI tools from the same organization.

DeepSeek-Reasonix

DeveloperApplication

DeepSeek-Reasonix - DeepSeek Terminal Coding Agent

AI Models by DeepSeek

Large language models from the same organization.

Model	Context Window	Price (In / Out per M)	Capabilities
DeepSeek V3.2 SpecialeCurrent	164K	$0.40 / $1.20	textcode
DeepSeek V3.2Current	164K	$0.26 / $0.42	textcode
DeepSeek V3.2 ExpCurrent	164K	$0.27 / $0.41	textcode
DeepSeek V3.1 TerminusCurrent	164K	$0.21 / $0.79	textcode
DeepSeek V4 Flash 0731Deprecated	1.0M	Deprecated	reasoningcodingtool usestructured output+2