Deepseek V3

deepseek-v3

The flagship model from the DeepSeek series, built for efficient reasoning, multilingual understanding.

Intelligence

Medium

Speed

Medium

Price

$0.20 • $0.58

Input • Output

Cache Price

$0.20 • $0.20

Read • Write

Input

Text

Output

Text

Start using Deepseek V3 now

Support Subscription & Pay-as-you-go matching your needs in Studio & API

View Pricing

Specifications

Context Window

128K

Max Output Tokens

Announced Date

2024-12-25

Released Date

2024-12-25

Modalities

Text

Input and output

Image

Not supported

Audio

Not supported

Video

Not supported

Endpoints

OpenAI Compatible

Chat Completions

/v1/chat/completions

Responses

/v1/responses

Embeddings

/v1/embeddings

Image Generation

/v1/images/generations

Image Edit

/v1/images/edits

Videos

/v1/videos

Claude Compatible

Messages

/v1/messages

Gemini Compatible

Generate Content

/v1beta/models/{model}:{operator}

Performance Rankings

View Full Leaderboard

Code Arena

Claude Opus 4.6

1,958

Gemini 3.1 Pro

1,797

GLM-5

1,614

Claude Opus 4.5

1,563

Gemini 3 Pro

1,557

Qwen3 30B A3B

Qwen3-Coder 480B A35B Instruct

Deepseek V3This Model

Chat Arena

Claude Opus 4.6

1,537

Grok-4 Fast Reasoning

1,347

GLM-4.5

1,342

Claude Sonnet 4.5

1,318

DeepSeek-V3.2-Exp

1,300

GPQA

Gemini 3.1 Pro

94.3%

GPT-5.2 Pro

93.2%

GPT-5.4

92.8%

GPT-5.2

92.4%

Gemini 3 Pro

91.9%

114

o1-mini

60.0%

115

Claude 3.5 Sonnet

59.4%

116

DeepSeek-V3This Model

59.1%

117

Gemini 1.5 Pro

59.1%

118

DeepSeek R1 Distill Qwen 14B

59.1%

AIME 2025

Grok-4 Heavy

100.0%

GPT-5.2

100.0%

Gemini 3 Pro

100.0%

Kimi K2-Thinking-0905

100.0%

GPT-5.2 Pro

100.0%

SWE-Bench

Claude Opus 4.5

80.9%

Claude Opus 4.6

80.8%

Gemini 3.1 Pro

80.6%

MiniMax M2.5

80.2%

GPT-5.2

80.0%

Claude 3.5 Sonnet

49.0%

DeepSeek-R1-0528

44.6%

DeepSeek-V3This Model

42.0%

o1-preview

41.3%

41.0%

ARC-AGI v2

Gemini 3.1 Pro

77.1%

GPT-5.4

73.3%

Claude Opus 4.6

68.8%

Claude Sonnet 4.6

58.3%

GPT-5.2 Pro

54.2%

Metric Definitions

LLM

Code Arena: Average score across coding arenas based on human votes.
Chat Arena: Human preference score from blind comparisons.
GPQA: Graduate-level science questions requiring expert knowledge.
AIME 2025: Recent math competition problems.
SWE-Bench: Real GitHub issues requiring code changes.
ARC-AGI v2: Abstract reasoning problems.

Image

IMAGE GEN: Human preference score for text-to-image generation.
IMAGE EDIT: Human preference score for image editing and transformation.

Video

Text to Video: Human preference score for text-to-video generation.
Image to Video: Human preference score for image-to-video generation.
Video to Video: Human preference score for video editing capabilities.

TTS

TTS: Human preference score for text-to-speech quality.

STT

STT: Human preference score for transcription accuracy.

Ready to use Deepseek V3?

Subscription & Pay-as-you-go

Powerful API Access

High Cost-Effectiveness

Interactive Studio UI

View Pricing