DataLearner logo
DE

DeepSeek-V3

Chat modelDeepSeek VDeepSeek V3

DeepSeek-V3

Release date: 2024-12-26Updated: 2025-03-21Views: 1,607
Parameters
681B
Context length
128K
Multilingual
Supported
Reasoning ability
No data

DeepSeek-V3 is an AI model published by DeepSeek-AI, released on 2024-12-26, for Chat model, with 681B parameters, and 128K context length, requiring about 687.9 GB storage, with a 92.30 score on BBH.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

DeepSeek-V3

Model basics

Reasoning traces
No data
Thinking modes
Thinking modes not supported
Context length
128K tokens
Max output length
No data
Model type
Chat model
Modality (in / out)
No data
Release date
2024-12-26
Model file size
687.9 GB
MoE architecture
No
Total params / Active params
681B / Not applicable
Knowledge cutoff
No data
DeepSeek-V3

Open source & experience

Code license
Weights license
DeepSeek License Agreement- Commercial use permitted
Live demo
N/A
DeepSeek-V3

Official resources

Paper
DataLearnerAI blog
DeepSeek-V3

API details

API speed
No data
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Standard
TypeConditionInputOutput
Text-$0.270/ 1M tokens$1.10/ 1M tokens
Cache PricingPrompt Cache
TypeTTLWriteRead
Text-$0.270/ 1M tokens$0.070/ 1M tokens
Text-$0.270/ 1M tokens
Cache = write
Text-$0.070/ 1M tokens
Cache = hit

“—” means the modality is not billed in that direction, or the vendor has not published a price for it.

DeepSeek-V3

Benchmark Results

DeepSeek-V3 currently shows benchmark results led by HumanEval (18 / 140, score 89), BBH (3 / 21, score 92.30), MMLU (18 / 124, score 88.50). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking

General Knowledge

4 evaluations
Benchmark / mode
Score
Rank/total
BBH
Standard Mode
92.30
3 / 21
MMLU
Standard Mode
88.50
18 / 124
MMLU Pro
Standard Mode
75.90
84 / 134
GPQA
Standard Mode
59.10
8 / 17

Coding and Software Engineer

2 evaluations
Benchmark / mode
Score
Rank/total
HumanEval
Standard Mode
89
18 / 140
LiveCodeBench
Standard Mode
34.60
115 / 128

Math and Reasoning

5 evaluations
Benchmark / mode
Score
Rank/total
MATH
Standard Mode
87.80
7 / 42
MATH-500
Standard Mode
87.80
40 / 45
AIME 2024
Standard Mode
39
52 / 62
4.30
16 / 19
FrontierMath
Standard Mode
1.70
49 / 60

General Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
GPQA Diamond
Standard Mode
59.10
231 / 274

Common Sense

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleQA
Standard Mode
24.90
31 / 47

Common Sense Reasoning

1 evaluations
Benchmark / mode
Score
Rank/total
SimpleBench
Standard Mode
18.90
90 / 94

Agent Level Benchmark

1 evaluations
Benchmark / mode
Score
Rank/total
Aider-Polyglot
Standard Mode
48.40
34 / 59
DeepSeek-V3

Publisher

DeepSeek-V3

Model Overview

DeepSeek-V3 is an AI model published by DeepSeek-AI, released on 2024-12-26, for Chat model, with 681B parameters, and 128K context length, requiring about 687.9 GB storage, with a 92.30 score on BBH.

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code