AIbase
EN
Home
AI Models Finder
Deepseek V3
D

Deepseek V3

DeepSeek-V3 is a large language model developed by the Chinese company DeepSeek. It adopts the MoE architecture, with 37 billion active parameters and a total of 671 billion parameters. It focuses on mathematics, programming, and Chinese language tasks, and its performance is comparable to that of top models such as GPT-4o.
Intelligence(Relatively Weak)
Speed(Relatively Slow)
Input Supported Modalities
No
Is Reasoning Model
128,000
Context Window
8,000
Maximum Output Tokens
-
Knowledge Cutoff
Go Compare

Pricing

￥1.94 /M tokens
Input
￥7.92 /M tokens
Output
￥4.93 /M tokens
Blended Price

Quick Simple Comparison

Input

Output

DeepSeek-Coder-V2
DeepSeek V3 0324 (Mar' 25)
DeepSeek V3 0324 (Mar '25)

Basic Parameters

DeepSeek-V3Technical Parameters
Parameter Count
671,000.0M
Context Length
128.00k tokens
Training Data Cutoff
Open Source Category
Open Source
Multimodal Support
Text Only
Throughput
14.80k
Release Date
2024-12-26
Response Speed
87.5 tokens/s

Benchmark Scores

Below is the performance of DeepSeek-V3 in various standard benchmark tests. These tests evaluate the model's capabilities in different tasks and domains.
Intelligence Index
40
Large Language Model Intelligence Level
Coding Index
42.6
Indicator of AI model performance on coding tasks
Math Index
80.2
Capability indicator in solving mathematical problems, mathematical reasoning, or performing math-related tasks
MMLU Pro
81.2
Massive Multitask Multimodal Understanding - Testing understanding of text, images, audio, and video
GPQA
68.4
Graduate Physics Questions Assessment - Testing advanced physics knowledge with diamond science-level questions
HLE
-
The model's comprehensive average score on the Hugging Face Open LLM Leaderboard
LiveCodeBench
49.2
Specific evaluation focused on assessing large language models' ability in real-world code writing and solving programming competition problems
SciCode
-
The model's capability in code generation for scientific computing or specific scientific domains
HumanEval
82.6
Score achieved by the AI model on the specific HumanEval benchmark test set
Math 500 Score
90.2
Score on the first 500 larger, more well-known mathematical benchmark tests
AIME Score
59.4
An indicator measuring an AI model's ability to solve high-difficulty mathematical competition problems (specifically AIME level)
AIbase
Empowering the Future, Your AI Solution Knowledge Base
English简体中文繁體中文にほんご
© 2025AIbase